├── models ├── __init__.py ├── .DS_Store ├── __pycache__ │ ├── common.cpython-36.pyc │ ├── common.cpython-37.pyc │ ├── yolo.cpython-37.pyc │ ├── __init__.cpython-36.pyc │ ├── __init__.cpython-37.pyc │ ├── experimental.cpython-36.pyc │ └── experimental.cpython-37.pyc ├── hub │ ├── yolov3-tiny.yaml │ ├── yolov5-fpn.yaml │ ├── yolov5s-transformer.yaml │ ├── yolov5-panet.yaml │ ├── yolov3.yaml │ ├── yolov3-spp.yaml │ ├── yolov5-p2.yaml │ ├── yolov5-p6.yaml │ ├── yolov5l6.yaml │ ├── yolov5m6.yaml │ ├── yolov5s6.yaml │ ├── yolov5x6.yaml │ ├── yolov5-p7.yaml │ └── anchors.yaml ├── yolov5l.yaml ├── yolov5m.yaml ├── yolov5s.yaml ├── yolov5x.yaml ├── experimental.py └── export.py ├── utils ├── __init__.py ├── aws │ ├── __init__.py │ ├── mime.sh │ ├── resume.py │ └── userdata.sh ├── wandb_logging │ ├── __init__.py │ └── log_dataset.py ├── .DS_Store ├── __pycache__ │ ├── general.cpython-37.pyc │ ├── metrics.cpython-37.pyc │ ├── plots.cpython-37.pyc │ ├── __init__.cpython-37.pyc │ ├── datasets.cpython-37.pyc │ ├── BaseDetector.cpython-37.pyc │ ├── autoanchor.cpython-37.pyc │ ├── google_utils.cpython-37.pyc │ └── torch_utils.cpython-37.pyc ├── google_app_engine │ ├── additional_requirements.txt │ ├── app.yaml │ └── Dockerfile ├── flask_rest_api │ ├── example_request.py │ ├── restapi.py │ └── README.md ├── BaseDetector.py ├── activations.py ├── google_utils.py ├── autoanchor.py ├── metrics.py └── loss.py ├── deep_sort ├── utils │ ├── __init__.py │ ├── __pycache__ │ │ ├── parser.cpython-36.pyc │ │ ├── parser.cpython-37.pyc │ │ ├── __init__.cpython-36.pyc │ │ └── __init__.cpython-37.pyc │ ├── asserts.py │ ├── log.py │ ├── tools.py │ ├── parser.py │ ├── draw.py │ ├── evaluation.py │ ├── io.py │ └── json_logger.py ├── deep_sort │ ├── deep │ │ ├── __init__.py │ │ ├── checkpoint │ │ │ └── .gitkeep │ │ ├── train.jpg │ │ ├── __pycache__ │ │ │ ├── model.cpython-36.pyc │ │ │ ├── model.cpython-37.pyc │ │ │ ├── __init__.cpython-36.pyc │ │ │ ├── __init__.cpython-37.pyc │ │ │ ├── feature_extractor.cpython-36.pyc │ │ │ └── feature_extractor.cpython-37.pyc │ │ ├── evaluate.py │ │ ├── feature_extractor.py │ │ ├── test.py │ │ ├── model.py │ │ ├── original_model.py │ │ └── train.py │ ├── sort │ │ ├── __init__.py │ │ ├── __pycache__ │ │ │ ├── track.cpython-36.pyc │ │ │ ├── track.cpython-37.pyc │ │ │ ├── __init__.cpython-36.pyc │ │ │ ├── __init__.cpython-37.pyc │ │ │ ├── tracker.cpython-36.pyc │ │ │ ├── tracker.cpython-37.pyc │ │ │ ├── detection.cpython-36.pyc │ │ │ ├── detection.cpython-37.pyc │ │ │ ├── iou_matching.cpython-36.pyc │ │ │ ├── iou_matching.cpython-37.pyc │ │ │ ├── kalman_filter.cpython-36.pyc │ │ │ ├── kalman_filter.cpython-37.pyc │ │ │ ├── nn_matching.cpython-36.pyc │ │ │ ├── nn_matching.cpython-37.pyc │ │ │ ├── preprocessing.cpython-36.pyc │ │ │ ├── preprocessing.cpython-37.pyc │ │ │ ├── linear_assignment.cpython-36.pyc │ │ │ └── linear_assignment.cpython-37.pyc │ │ ├── detection.py │ │ ├── preprocessing.py │ │ ├── iou_matching.py │ │ ├── tracker.py │ │ ├── track.py │ │ ├── nn_matching.py │ │ ├── linear_assignment.py │ │ └── kalman_filter.py │ ├── README.md │ ├── __pycache__ │ │ ├── __init__.cpython-36.pyc │ │ ├── __init__.cpython-37.pyc │ │ ├── deep_sort.cpython-36.pyc │ │ └── deep_sort.cpython-37.pyc │ ├── __init__.py │ └── deep_sort.py ├── .DS_Store └── configs │ └── deep_sort.yaml ├── .DS_Store ├── image.jpg ├── requirements.txt ├── demoY.py ├── demoX.py ├── AIDetector_pytorch.py ├── tracker.py └── README.md /models/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /utils/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /utils/aws/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /deep_sort/utils/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /utils/wandb_logging/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/checkpoint/.gitkeep: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /.DS_Store: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/.DS_Store -------------------------------------------------------------------------------- /image.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/image.jpg -------------------------------------------------------------------------------- /utils/.DS_Store: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/.DS_Store -------------------------------------------------------------------------------- /models/.DS_Store: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/.DS_Store -------------------------------------------------------------------------------- /deep_sort/.DS_Store: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/.DS_Store -------------------------------------------------------------------------------- /deep_sort/deep_sort/README.md: -------------------------------------------------------------------------------- 1 | # Deep Sort 2 | 3 | This is the implemention of deep sort with pytorch. -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/train.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/train.jpg -------------------------------------------------------------------------------- /models/__pycache__/common.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/common.cpython-36.pyc -------------------------------------------------------------------------------- /models/__pycache__/common.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/common.cpython-37.pyc -------------------------------------------------------------------------------- /models/__pycache__/yolo.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/yolo.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/general.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/general.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/metrics.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/metrics.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/plots.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/plots.cpython-37.pyc -------------------------------------------------------------------------------- /models/__pycache__/__init__.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/__init__.cpython-36.pyc -------------------------------------------------------------------------------- /models/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/datasets.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/datasets.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/BaseDetector.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/BaseDetector.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/autoanchor.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/autoanchor.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/google_utils.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/google_utils.cpython-37.pyc -------------------------------------------------------------------------------- /utils/__pycache__/torch_utils.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/utils/__pycache__/torch_utils.cpython-37.pyc -------------------------------------------------------------------------------- /models/__pycache__/experimental.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/experimental.cpython-36.pyc -------------------------------------------------------------------------------- /models/__pycache__/experimental.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/models/__pycache__/experimental.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/utils/__pycache__/parser.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/utils/__pycache__/parser.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/utils/__pycache__/parser.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/utils/__pycache__/parser.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/utils/__pycache__/__init__.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/utils/__pycache__/__init__.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/utils/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/utils/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/__pycache__/__init__.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/__pycache__/__init__.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/__pycache__/deep_sort.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/__pycache__/deep_sort.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/__pycache__/deep_sort.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/__pycache__/deep_sort.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__pycache__/model.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/__pycache__/model.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__pycache__/model.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/__pycache__/model.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/track.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/track.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/track.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/track.cpython-37.pyc -------------------------------------------------------------------------------- /utils/google_app_engine/additional_requirements.txt: -------------------------------------------------------------------------------- 1 | # add these requirements in your app on top of the existing ones 2 | pip==18.1 3 | Flask==1.0.2 4 | gunicorn==19.9.0 5 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__pycache__/__init__.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/__pycache__/__init__.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/__init__.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/__init__.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/__init__.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/__init__.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/tracker.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/tracker.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/tracker.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/tracker.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/detection.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/detection.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/detection.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/detection.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/iou_matching.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/iou_matching.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/iou_matching.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/iou_matching.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/kalman_filter.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/kalman_filter.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/kalman_filter.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/kalman_filter.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/nn_matching.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/nn_matching.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/nn_matching.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/nn_matching.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/preprocessing.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/preprocessing.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/preprocessing.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/preprocessing.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__pycache__/feature_extractor.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/__pycache__/feature_extractor.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/__pycache__/feature_extractor.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/deep/__pycache__/feature_extractor.cpython-37.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/linear_assignment.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/linear_assignment.cpython-36.pyc -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/__pycache__/linear_assignment.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Whiffe/yolov5-deepsort/HEAD/deep_sort/deep_sort/sort/__pycache__/linear_assignment.cpython-37.pyc -------------------------------------------------------------------------------- /utils/google_app_engine/app.yaml: -------------------------------------------------------------------------------- 1 | runtime: custom 2 | env: flex 3 | 4 | service: yolov5app 5 | 6 | liveness_check: 7 | initial_delay_sec: 600 8 | 9 | manual_scaling: 10 | instances: 1 11 | resources: 12 | cpu: 1 13 | memory_gb: 4 14 | disk_size_gb: 20 -------------------------------------------------------------------------------- /deep_sort/configs/deep_sort.yaml: -------------------------------------------------------------------------------- 1 | DEEPSORT: 2 | REID_CKPT: "deep_sort/deep_sort/deep/checkpoint/ckpt.t7" 3 | MAX_DIST: 0.2 4 | MIN_CONFIDENCE: 0.3 5 | NMS_MAX_OVERLAP: 0.5 6 | MAX_IOU_DISTANCE: 0.7 7 | MAX_AGE: 70 8 | N_INIT: 3 9 | NN_BUDGET: 100 10 | 11 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/evaluate.py: -------------------------------------------------------------------------------- 1 | import torch 2 | 3 | features = torch.load("features.pth") 4 | qf = features["qf"] 5 | ql = features["ql"] 6 | gf = features["gf"] 7 | gl = features["gl"] 8 | 9 | scores = qf.mm(gf.t()) 10 | res = scores.topk(5, dim=1)[1][:,0] 11 | top1correct = gl[res].eq(ql).sum().item() 12 | 13 | print("Acc top1:{:.3f}".format(top1correct/ql.size(0))) 14 | 15 | 16 | -------------------------------------------------------------------------------- /deep_sort/utils/asserts.py: -------------------------------------------------------------------------------- 1 | from os import environ 2 | 3 | 4 | def assert_in(file, files_to_check): 5 | if file not in files_to_check: 6 | raise AssertionError("{} does not exist in the list".format(str(file))) 7 | return True 8 | 9 | 10 | def assert_in_env(check_list: list): 11 | for item in check_list: 12 | assert_in(item, environ.keys()) 13 | return True 14 | -------------------------------------------------------------------------------- /utils/flask_rest_api/example_request.py: -------------------------------------------------------------------------------- 1 | """Perform test request""" 2 | import pprint 3 | 4 | import requests 5 | 6 | DETECTION_URL = "http://localhost:5000/v1/object-detection/yolov5s" 7 | TEST_IMAGE = "zidane.jpg" 8 | 9 | image_data = open(TEST_IMAGE, "rb").read() 10 | 11 | response = requests.post(DETECTION_URL, files={"image": image_data}).json() 12 | 13 | pprint.pprint(response) 14 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | # pip install -r requirements.txt 2 | 3 | # base ---------------------------------------- 4 | 5 | matplotlib>=3.2.2 6 | numpy>=1.18.5 7 | opencv-python>=4.1.2 8 | Pillow 9 | PyYAML>=5.3.1 10 | scipy>=1.4.1 11 | torch>=1.7.0 12 | torchvision>=0.8.1 13 | tqdm>=4.41.0 14 | 15 | # plotting ------------------------------------ 16 | 17 | seaborn>=0.11.0 18 | pandas 19 | imutils 20 | 21 | # deep_sort ----------------------------------- 22 | 23 | easydict 24 | -------------------------------------------------------------------------------- /deep_sort/utils/log.py: -------------------------------------------------------------------------------- 1 | import logging 2 | 3 | 4 | def get_logger(name='root'): 5 | formatter = logging.Formatter( 6 | # fmt='%(asctime)s [%(levelname)s]: %(filename)s(%(funcName)s:%(lineno)s) >> %(message)s') 7 | fmt='%(asctime)s [%(levelname)s]: %(message)s', datefmt='%Y-%m-%d %H:%M:%S') 8 | 9 | handler = logging.StreamHandler() 10 | handler.setFormatter(formatter) 11 | 12 | logger = logging.getLogger(name) 13 | logger.setLevel(logging.INFO) 14 | logger.addHandler(handler) 15 | return logger 16 | 17 | 18 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/__init__.py: -------------------------------------------------------------------------------- 1 | from .deep_sort import DeepSort 2 | 3 | 4 | __all__ = ['DeepSort', 'build_tracker'] 5 | 6 | 7 | def build_tracker(cfg, use_cuda): 8 | return DeepSort(cfg.DEEPSORT.REID_CKPT, 9 | max_dist=cfg.DEEPSORT.MAX_DIST, min_confidence=cfg.DEEPSORT.MIN_CONFIDENCE, 10 | nms_max_overlap=cfg.DEEPSORT.NMS_MAX_OVERLAP, max_iou_distance=cfg.DEEPSORT.MAX_IOU_DISTANCE, 11 | max_age=cfg.DEEPSORT.MAX_AGE, n_init=cfg.DEEPSORT.N_INIT, nn_budget=cfg.DEEPSORT.NN_BUDGET, use_cuda=use_cuda) 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | -------------------------------------------------------------------------------- /utils/aws/mime.sh: -------------------------------------------------------------------------------- 1 | # AWS EC2 instance startup 'MIME' script https://aws.amazon.com/premiumsupport/knowledge-center/execute-user-data-ec2/ 2 | # This script will run on every instance restart, not only on first start 3 | # --- DO NOT COPY ABOVE COMMENTS WHEN PASTING INTO USERDATA --- 4 | 5 | Content-Type: multipart/mixed; boundary="//" 6 | MIME-Version: 1.0 7 | 8 | --// 9 | Content-Type: text/cloud-config; charset="us-ascii" 10 | MIME-Version: 1.0 11 | Content-Transfer-Encoding: 7bit 12 | Content-Disposition: attachment; filename="cloud-config.txt" 13 | 14 | #cloud-config 15 | cloud_final_modules: 16 | - [scripts-user, always] 17 | 18 | --// 19 | Content-Type: text/x-shellscript; charset="us-ascii" 20 | MIME-Version: 1.0 21 | Content-Transfer-Encoding: 7bit 22 | Content-Disposition: attachment; filename="userdata.txt" 23 | 24 | #!/bin/bash 25 | # --- paste contents of userdata.sh here --- 26 | --// 27 | -------------------------------------------------------------------------------- /deep_sort/utils/tools.py: -------------------------------------------------------------------------------- 1 | from functools import wraps 2 | from time import time 3 | 4 | 5 | def is_video(ext: str): 6 | """ 7 | Returns true if ext exists in 8 | allowed_exts for video files. 9 | 10 | Args: 11 | ext: 12 | 13 | Returns: 14 | 15 | """ 16 | 17 | allowed_exts = ('.mp4', '.webm', '.ogg', '.avi', '.wmv', '.mkv', '.3gp') 18 | return any((ext.endswith(x) for x in allowed_exts)) 19 | 20 | 21 | def tik_tok(func): 22 | """ 23 | keep track of time for each process. 24 | Args: 25 | func: 26 | 27 | Returns: 28 | 29 | """ 30 | @wraps(func) 31 | def _time_it(*args, **kwargs): 32 | start = time() 33 | try: 34 | return func(*args, **kwargs) 35 | finally: 36 | end_ = time() 37 | print("time: {:.03f}s, fps: {:.03f}".format(end_ - start, 1 / (end_ - start))) 38 | 39 | return _time_it 40 | -------------------------------------------------------------------------------- /utils/wandb_logging/log_dataset.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | 3 | import yaml 4 | 5 | from wandb_utils import WandbLogger 6 | 7 | WANDB_ARTIFACT_PREFIX = 'wandb-artifact://' 8 | 9 | 10 | def create_dataset_artifact(opt): 11 | with open(opt.data) as f: 12 | data = yaml.safe_load(f) # data dict 13 | logger = WandbLogger(opt, '', None, data, job_type='Dataset Creation') 14 | 15 | 16 | if __name__ == '__main__': 17 | parser = argparse.ArgumentParser() 18 | parser.add_argument('--data', type=str, default='data/coco128.yaml', help='data.yaml path') 19 | parser.add_argument('--single-cls', action='store_true', help='train as single-class dataset') 20 | parser.add_argument('--project', type=str, default='YOLOv5', help='name of W&B Project') 21 | opt = parser.parse_args() 22 | opt.resume = False # Explicitly disallow resume check for dataset upload job 23 | 24 | create_dataset_artifact(opt) 25 | -------------------------------------------------------------------------------- /utils/google_app_engine/Dockerfile: -------------------------------------------------------------------------------- 1 | FROM gcr.io/google-appengine/python 2 | 3 | # Create a virtualenv for dependencies. This isolates these packages from 4 | # system-level packages. 5 | # Use -p python3 or -p python3.7 to select python version. Default is version 2. 6 | RUN virtualenv /env -p python3 7 | 8 | # Setting these environment variables are the same as running 9 | # source /env/bin/activate. 10 | ENV VIRTUAL_ENV /env 11 | ENV PATH /env/bin:$PATH 12 | 13 | RUN apt-get update && apt-get install -y python-opencv 14 | 15 | # Copy the application's requirements.txt and run pip to install all 16 | # dependencies into the virtualenv. 17 | ADD requirements.txt /app/requirements.txt 18 | RUN pip install -r /app/requirements.txt 19 | 20 | # Add the application source code. 21 | ADD . /app 22 | 23 | # Run a WSGI server to serve the application. gunicorn must be declared as 24 | # a dependency in requirements.txt. 25 | CMD gunicorn -b :$PORT main:app 26 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/detection.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | import numpy as np 3 | 4 | 5 | class Detection(object): 6 | 7 | def __init__(self, tlwh, cls_, confidence, feature): 8 | self.tlwh = np.asarray(tlwh, dtype=np.float) 9 | self.cls_ = cls_ 10 | self.confidence = float(confidence) 11 | self.feature = np.asarray(feature, dtype=np.float32) 12 | 13 | def to_tlbr(self): 14 | """Convert bounding box to format `(min x, min y, max x, max y)`, i.e., 15 | `(top left, bottom right)`. 16 | """ 17 | ret = self.tlwh.copy() 18 | ret[2:] += ret[:2] 19 | return ret 20 | 21 | def to_xyah(self): 22 | """Convert bounding box to format `(center x, center y, aspect ratio, 23 | height)`, where the aspect ratio is `width / height`. 24 | """ 25 | ret = self.tlwh.copy() 26 | ret[:2] += ret[2:] / 2 27 | ret[2] /= ret[3] 28 | return ret 29 | -------------------------------------------------------------------------------- /demoY.py: -------------------------------------------------------------------------------- 1 | from AIDetector_pytorch import Detector 2 | import imutils 3 | import cv2 4 | 5 | def main(): 6 | 7 | name = 'demo' 8 | 9 | det = Detector() 10 | cap = cv2.VideoCapture('./test.mp4') 11 | fps = int(cap.get(5)) 12 | print('fps:', fps) 13 | t = int(1000/fps) 14 | 15 | videoWriter = None 16 | 17 | while True: 18 | 19 | # try: 20 | _, im = cap.read() 21 | if im is None: 22 | break 23 | 24 | result = det.feedCap(im) 25 | result = result['frame'] 26 | result = imutils.resize(result, height=500) 27 | if videoWriter is None: 28 | fourcc = cv2.VideoWriter_fourcc( 29 | 'm', 'p', '4', 'v') # opencv3.0 30 | videoWriter = cv2.VideoWriter( 31 | 'result.mp4', fourcc, fps, (result.shape[1], result.shape[0])) 32 | 33 | videoWriter.write(result) 34 | 35 | cap.release() 36 | videoWriter.release() 37 | cv2.destroyAllWindows() 38 | 39 | if __name__ == '__main__': 40 | 41 | main() -------------------------------------------------------------------------------- /deep_sort/utils/parser.py: -------------------------------------------------------------------------------- 1 | import os 2 | import yaml 3 | from easydict import EasyDict as edict 4 | 5 | class YamlParser(edict): 6 | """ 7 | This is yaml parser based on EasyDict. 8 | """ 9 | def __init__(self, cfg_dict=None, config_file=None): 10 | if cfg_dict is None: 11 | cfg_dict = {} 12 | 13 | if config_file is not None: 14 | assert(os.path.isfile(config_file)) 15 | with open(config_file, 'r') as fo: 16 | cfg_dict.update(yaml.load(fo.read())) 17 | 18 | super(YamlParser, self).__init__(cfg_dict) 19 | 20 | 21 | def merge_from_file(self, config_file): 22 | with open(config_file, 'r') as fo: 23 | self.update(yaml.load(fo.read())) 24 | 25 | 26 | def merge_from_dict(self, config_dict): 27 | self.update(config_dict) 28 | 29 | 30 | def get_config(config_file=None): 31 | return YamlParser(config_file=config_file) 32 | 33 | 34 | if __name__ == "__main__": 35 | cfg = YamlParser(config_file="../configs/yolov3.yaml") 36 | cfg.merge_from_file("../configs/deep_sort.yaml") 37 | 38 | import ipdb; ipdb.set_trace() -------------------------------------------------------------------------------- /utils/flask_rest_api/restapi.py: -------------------------------------------------------------------------------- 1 | """ 2 | Run a rest API exposing the yolov5s object detection model 3 | """ 4 | import argparse 5 | import io 6 | 7 | import torch 8 | from PIL import Image 9 | from flask import Flask, request 10 | 11 | app = Flask(__name__) 12 | 13 | DETECTION_URL = "/v1/object-detection/yolov5s" 14 | 15 | 16 | @app.route(DETECTION_URL, methods=["POST"]) 17 | def predict(): 18 | if not request.method == "POST": 19 | return 20 | 21 | if request.files.get("image"): 22 | image_file = request.files["image"] 23 | image_bytes = image_file.read() 24 | 25 | img = Image.open(io.BytesIO(image_bytes)) 26 | 27 | results = model(img, size=640) # reduce size=320 for faster inference 28 | return results.pandas().xyxy[0].to_json(orient="records") 29 | 30 | 31 | if __name__ == "__main__": 32 | parser = argparse.ArgumentParser(description="Flask API exposing YOLOv5 model") 33 | parser.add_argument("--port", default=5000, type=int, help="port number") 34 | args = parser.parse_args() 35 | 36 | model = torch.hub.load("ultralytics/yolov5", "yolov5s", force_reload=True) # force_reload to recache 37 | app.run(host="0.0.0.0", port=args.port) # debug=True causes Restarting with stat 38 | -------------------------------------------------------------------------------- /utils/aws/resume.py: -------------------------------------------------------------------------------- 1 | # Resume all interrupted trainings in yolov5/ dir including DDP trainings 2 | # Usage: $ python utils/aws/resume.py 3 | 4 | import os 5 | import sys 6 | from pathlib import Path 7 | 8 | import torch 9 | import yaml 10 | 11 | sys.path.append('./') # to run '$ python *.py' files in subdirectories 12 | 13 | port = 0 # --master_port 14 | path = Path('').resolve() 15 | for last in path.rglob('*/**/last.pt'): 16 | ckpt = torch.load(last) 17 | if ckpt['optimizer'] is None: 18 | continue 19 | 20 | # Load opt.yaml 21 | with open(last.parent.parent / 'opt.yaml') as f: 22 | opt = yaml.safe_load(f) 23 | 24 | # Get device count 25 | d = opt['device'].split(',') # devices 26 | nd = len(d) # number of devices 27 | ddp = nd > 1 or (nd == 0 and torch.cuda.device_count() > 1) # distributed data parallel 28 | 29 | if ddp: # multi-GPU 30 | port += 1 31 | cmd = f'python -m torch.distributed.launch --nproc_per_node {nd} --master_port {port} train.py --resume {last}' 32 | else: # single-GPU 33 | cmd = f'python train.py --resume {last}' 34 | 35 | cmd += ' > /dev/null 2>&1 &' # redirect output to dev/null and run in daemon thread 36 | print(cmd) 37 | os.system(cmd) 38 | -------------------------------------------------------------------------------- /deep_sort/utils/draw.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import cv2 3 | 4 | palette = (2 ** 11 - 1, 2 ** 15 - 1, 2 ** 20 - 1) 5 | 6 | 7 | def compute_color_for_labels(label): 8 | """ 9 | Simple function that adds fixed color depending on the class 10 | """ 11 | color = [int((p * (label ** 2 - label + 1)) % 255) for p in palette] 12 | return tuple(color) 13 | 14 | 15 | def draw_boxes(img, bbox, identities=None, offset=(0,0)): 16 | for i,box in enumerate(bbox): 17 | x1,y1,x2,y2 = [int(i) for i in box] 18 | x1 += offset[0] 19 | x2 += offset[0] 20 | y1 += offset[1] 21 | y2 += offset[1] 22 | # box text and bar 23 | id = int(identities[i]) if identities is not None else 0 24 | color = compute_color_for_labels(id) 25 | label = '{}{:d}'.format("", id) 26 | t_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_PLAIN, 2 , 2)[0] 27 | cv2.rectangle(img,(x1, y1),(x2,y2),color,3) 28 | cv2.rectangle(img,(x1, y1),(x1+t_size[0]+3,y1+t_size[1]+4), color,-1) 29 | cv2.putText(img,label,(x1,y1+t_size[1]+4), cv2.FONT_HERSHEY_PLAIN, 2, [255,255,255], 2) 30 | return img 31 | 32 | 33 | 34 | if __name__ == '__main__': 35 | for i in range(82): 36 | print(compute_color_for_labels(i)) 37 | -------------------------------------------------------------------------------- /utils/BaseDetector.py: -------------------------------------------------------------------------------- 1 | from tracker import update_tracker 2 | import cv2 3 | 4 | 5 | class baseDet(object): 6 | 7 | def __init__(self): 8 | 9 | self.img_size = 640 10 | self.threshold = 0.3 11 | self.stride = 1 12 | 13 | def build_config(self): 14 | 15 | self.faceTracker = {} 16 | self.faceClasses = {} 17 | self.faceLocation1 = {} 18 | self.faceLocation2 = {} 19 | self.frameCounter = 0 20 | self.currentCarID = 0 21 | self.recorded = [] 22 | 23 | self.font = cv2.FONT_HERSHEY_SIMPLEX 24 | 25 | def feedCap(self, im): 26 | 27 | retDict = { 28 | 'frame': None, 29 | 'faces': None, 30 | 'list_of_ids': None, 31 | 'face_bboxes': [] 32 | } 33 | self.frameCounter += 1 34 | 35 | im, faces, face_bboxes = update_tracker(self, im) 36 | 37 | retDict['frame'] = im 38 | retDict['faces'] = faces 39 | retDict['face_bboxes'] = face_bboxes 40 | 41 | return retDict 42 | 43 | def init_model(self): 44 | raise EOFError("Undefined model type.") 45 | 46 | def preprocess(self): 47 | raise EOFError("Undefined model type.") 48 | 49 | def detect(self): 50 | raise EOFError("Undefined model type.") 51 | -------------------------------------------------------------------------------- /demoX.py: -------------------------------------------------------------------------------- 1 | from AIDetector_pytorch import Detector 2 | import imutils 3 | import cv2 4 | 5 | def main(): 6 | 7 | name = 'demo' 8 | 9 | det = Detector() 10 | cap = cv2.VideoCapture('./test.mp4') 11 | fps = int(cap.get(5)) 12 | print('fps:', fps) 13 | t = int(1000/fps) 14 | 15 | videoWriter = None 16 | 17 | while True: 18 | 19 | # try: 20 | _, im = cap.read() 21 | if im is None: 22 | break 23 | 24 | result = det.feedCap(im) 25 | result = result['frame'] 26 | result = imutils.resize(result, height=500) 27 | if videoWriter is None: 28 | fourcc = cv2.VideoWriter_fourcc( 29 | 'm', 'p', '4', 'v') # opencv3.0 30 | videoWriter = cv2.VideoWriter( 31 | 'result.mp4', fourcc, fps, (result.shape[1], result.shape[0])) 32 | 33 | videoWriter.write(result) 34 | cv2.imshow(name, result) 35 | cv2.waitKey(t) 36 | 37 | if cv2.getWindowProperty(name, cv2.WND_PROP_AUTOSIZE) < 1: 38 | # 点x退出 39 | break 40 | # except Exception as e: 41 | # print(e) 42 | # break 43 | 44 | cap.release() 45 | videoWriter.release() 46 | cv2.destroyAllWindows() 47 | 48 | if __name__ == '__main__': 49 | 50 | main() -------------------------------------------------------------------------------- /utils/aws/userdata.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # AWS EC2 instance startup script https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html 3 | # This script will run only once on first instance start (for a re-start script see mime.sh) 4 | # /home/ubuntu (ubuntu) or /home/ec2-user (amazon-linux) is working dir 5 | # Use >300 GB SSD 6 | 7 | cd home/ubuntu 8 | if [ ! -d yolov5 ]; then 9 | echo "Running first-time script." # install dependencies, download COCO, pull Docker 10 | git clone https://github.com/ultralytics/yolov5 -b master && sudo chmod -R 777 yolov5 11 | cd yolov5 12 | bash data/scripts/get_coco.sh && echo "Data done." & 13 | sudo docker pull ultralytics/yolov5:latest && echo "Docker done." & 14 | python -m pip install --upgrade pip && pip install -r requirements.txt && python detect.py && echo "Requirements done." & 15 | wait && echo "All tasks done." # finish background tasks 16 | else 17 | echo "Running re-start script." # resume interrupted runs 18 | i=0 19 | list=$(sudo docker ps -qa) # container list i.e. $'one\ntwo\nthree\nfour' 20 | while IFS= read -r id; do 21 | ((i++)) 22 | echo "restarting container $i: $id" 23 | sudo docker start $id 24 | # sudo docker exec -it $id python train.py --resume # single-GPU 25 | sudo docker exec -d $id python utils/aws/resume.py # multi-scenario 26 | done <<<"$list" 27 | fi 28 | -------------------------------------------------------------------------------- /models/hub/yolov3-tiny.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,14, 23,27, 37,58] # P4/16 9 | - [81,82, 135,169, 344,319] # P5/32 10 | 11 | # YOLOv3-tiny backbone 12 | backbone: 13 | # [from, number, module, args] 14 | [[-1, 1, Conv, [16, 3, 1]], # 0 15 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 1-P1/2 16 | [-1, 1, Conv, [32, 3, 1]], 17 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 3-P2/4 18 | [-1, 1, Conv, [64, 3, 1]], 19 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 5-P3/8 20 | [-1, 1, Conv, [128, 3, 1]], 21 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 7-P4/16 22 | [-1, 1, Conv, [256, 3, 1]], 23 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 9-P5/32 24 | [-1, 1, Conv, [512, 3, 1]], 25 | [-1, 1, nn.ZeroPad2d, [[0, 1, 0, 1]]], # 11 26 | [-1, 1, nn.MaxPool2d, [2, 1, 0]], # 12 27 | ] 28 | 29 | # YOLOv3-tiny head 30 | head: 31 | [[-1, 1, Conv, [1024, 3, 1]], 32 | [-1, 1, Conv, [256, 1, 1]], 33 | [-1, 1, Conv, [512, 3, 1]], # 15 (P5/32-large) 34 | 35 | [-2, 1, Conv, [128, 1, 1]], 36 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 37 | [[-1, 8], 1, Concat, [1]], # cat backbone P4 38 | [-1, 1, Conv, [256, 3, 1]], # 19 (P4/16-medium) 39 | 40 | [[19, 15], 1, Detect, [nc, anchors]], # Detect(P4, P5) 41 | ] 42 | -------------------------------------------------------------------------------- /models/hub/yolov5-fpn.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, Bottleneck, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, BottleneckCSP, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, BottleneckCSP, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 6, BottleneckCSP, [1024]], # 9 25 | ] 26 | 27 | # YOLOv5 FPN head 28 | head: 29 | [[-1, 3, BottleneckCSP, [1024, False]], # 10 (P5/32-large) 30 | 31 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 32 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 33 | [-1, 1, Conv, [512, 1, 1]], 34 | [-1, 3, BottleneckCSP, [512, False]], # 14 (P4/16-medium) 35 | 36 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 37 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 38 | [-1, 1, Conv, [256, 1, 1]], 39 | [-1, 3, BottleneckCSP, [256, False]], # 18 (P3/8-small) 40 | 41 | [[18, 14, 10], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 42 | ] 43 | -------------------------------------------------------------------------------- /models/yolov5l.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /models/yolov5m.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.67 # model depth multiple 4 | width_multiple: 0.75 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /models/yolov5s.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.33 # model depth multiple 4 | width_multiple: 0.50 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /models/yolov5x.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.33 # model depth multiple 4 | width_multiple: 1.25 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /models/hub/yolov5s-transformer.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.33 # model depth multiple 4 | width_multiple: 0.50 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3TR, [1024, False]], # 9 <-------- C3TR() Transformer module 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /models/hub/yolov5-panet.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, BottleneckCSP, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, BottleneckCSP, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, BottleneckCSP, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, BottleneckCSP, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 PANet head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, BottleneckCSP, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, BottleneckCSP, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, BottleneckCSP, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, BottleneckCSP, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /models/hub/yolov3.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # darknet53 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Conv, [32, 3, 1]], # 0 16 | [-1, 1, Conv, [64, 3, 2]], # 1-P1/2 17 | [-1, 1, Bottleneck, [64]], 18 | [-1, 1, Conv, [128, 3, 2]], # 3-P2/4 19 | [-1, 2, Bottleneck, [128]], 20 | [-1, 1, Conv, [256, 3, 2]], # 5-P3/8 21 | [-1, 8, Bottleneck, [256]], 22 | [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 23 | [-1, 8, Bottleneck, [512]], 24 | [-1, 1, Conv, [1024, 3, 2]], # 9-P5/32 25 | [-1, 4, Bottleneck, [1024]], # 10 26 | ] 27 | 28 | # YOLOv3 head 29 | head: 30 | [[-1, 1, Bottleneck, [1024, False]], 31 | [-1, 1, Conv, [512, [1, 1]]], 32 | [-1, 1, Conv, [1024, 3, 1]], 33 | [-1, 1, Conv, [512, 1, 1]], 34 | [-1, 1, Conv, [1024, 3, 1]], # 15 (P5/32-large) 35 | 36 | [-2, 1, Conv, [256, 1, 1]], 37 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 38 | [[-1, 8], 1, Concat, [1]], # cat backbone P4 39 | [-1, 1, Bottleneck, [512, False]], 40 | [-1, 1, Bottleneck, [512, False]], 41 | [-1, 1, Conv, [256, 1, 1]], 42 | [-1, 1, Conv, [512, 3, 1]], # 22 (P4/16-medium) 43 | 44 | [-2, 1, Conv, [128, 1, 1]], 45 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 46 | [[-1, 6], 1, Concat, [1]], # cat backbone P3 47 | [-1, 1, Bottleneck, [256, False]], 48 | [-1, 2, Bottleneck, [256, False]], # 27 (P3/8-small) 49 | 50 | [[27, 22, 15], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 51 | ] 52 | -------------------------------------------------------------------------------- /models/hub/yolov3-spp.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # darknet53 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Conv, [32, 3, 1]], # 0 16 | [-1, 1, Conv, [64, 3, 2]], # 1-P1/2 17 | [-1, 1, Bottleneck, [64]], 18 | [-1, 1, Conv, [128, 3, 2]], # 3-P2/4 19 | [-1, 2, Bottleneck, [128]], 20 | [-1, 1, Conv, [256, 3, 2]], # 5-P3/8 21 | [-1, 8, Bottleneck, [256]], 22 | [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 23 | [-1, 8, Bottleneck, [512]], 24 | [-1, 1, Conv, [1024, 3, 2]], # 9-P5/32 25 | [-1, 4, Bottleneck, [1024]], # 10 26 | ] 27 | 28 | # YOLOv3-SPP head 29 | head: 30 | [[-1, 1, Bottleneck, [1024, False]], 31 | [-1, 1, SPP, [512, [5, 9, 13]]], 32 | [-1, 1, Conv, [1024, 3, 1]], 33 | [-1, 1, Conv, [512, 1, 1]], 34 | [-1, 1, Conv, [1024, 3, 1]], # 15 (P5/32-large) 35 | 36 | [-2, 1, Conv, [256, 1, 1]], 37 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 38 | [[-1, 8], 1, Concat, [1]], # cat backbone P4 39 | [-1, 1, Bottleneck, [512, False]], 40 | [-1, 1, Bottleneck, [512, False]], 41 | [-1, 1, Conv, [256, 1, 1]], 42 | [-1, 1, Conv, [512, 3, 1]], # 22 (P4/16-medium) 43 | 44 | [-2, 1, Conv, [128, 1, 1]], 45 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 46 | [[-1, 6], 1, Concat, [1]], # cat backbone P3 47 | [-1, 1, Bottleneck, [256, False]], 48 | [-1, 2, Bottleneck, [256, False]], # 27 (P3/8-small) 49 | 50 | [[27, 22, 15], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 51 | ] 52 | -------------------------------------------------------------------------------- /models/hub/yolov5-p2.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 3 8 | 9 | # YOLOv5 backbone 10 | backbone: 11 | # [from, number, module, args] 12 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 13 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 14 | [ -1, 3, C3, [ 128 ] ], 15 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 16 | [ -1, 9, C3, [ 256 ] ], 17 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 18 | [ -1, 9, C3, [ 512 ] ], 19 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 7-P5/32 20 | [ -1, 1, SPP, [ 1024, [ 5, 9, 13 ] ] ], 21 | [ -1, 3, C3, [ 1024, False ] ], # 9 22 | ] 23 | 24 | # YOLOv5 head 25 | head: 26 | [ [ -1, 1, Conv, [ 512, 1, 1 ] ], 27 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 28 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 29 | [ -1, 3, C3, [ 512, False ] ], # 13 30 | 31 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 32 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 33 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 34 | [ -1, 3, C3, [ 256, False ] ], # 17 (P3/8-small) 35 | 36 | [ -1, 1, Conv, [ 128, 1, 1 ] ], 37 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 38 | [ [ -1, 2 ], 1, Concat, [ 1 ] ], # cat backbone P2 39 | [ -1, 1, C3, [ 128, False ] ], # 21 (P2/4-xsmall) 40 | 41 | [ -1, 1, Conv, [ 128, 3, 2 ] ], 42 | [ [ -1, 18 ], 1, Concat, [ 1 ] ], # cat head P3 43 | [ -1, 3, C3, [ 256, False ] ], # 24 (P3/8-small) 44 | 45 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 46 | [ [ -1, 14 ], 1, Concat, [ 1 ] ], # cat head P4 47 | [ -1, 3, C3, [ 512, False ] ], # 27 (P4/16-medium) 48 | 49 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 50 | [ [ -1, 10 ], 1, Concat, [ 1 ] ], # cat head P5 51 | [ -1, 3, C3, [ 1024, False ] ], # 30 (P5/32-large) 52 | 53 | [ [ 24, 27, 30 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5) 54 | ] 55 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/feature_extractor.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torchvision.transforms as transforms 3 | import numpy as np 4 | import cv2 5 | import logging 6 | 7 | from .model import Net 8 | 9 | class Extractor(object): 10 | def __init__(self, model_path, use_cuda=True): 11 | self.net = Net(reid=True) 12 | self.device = "cuda" if torch.cuda.is_available() and use_cuda else "cpu" 13 | state_dict = torch.load(model_path, map_location=lambda storage, loc: storage)['net_dict'] 14 | self.net.load_state_dict(state_dict) 15 | logger = logging.getLogger("root.tracker") 16 | logger.info("Loading weights from {}... Done!".format(model_path)) 17 | self.net.to(self.device) 18 | self.size = (64, 128) 19 | self.norm = transforms.Compose([ 20 | transforms.ToTensor(), 21 | transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]), 22 | ]) 23 | 24 | 25 | 26 | def _preprocess(self, im_crops): 27 | """ 28 | TODO: 29 | 1. to float with scale from 0 to 1 30 | 2. resize to (64, 128) as Market1501 dataset did 31 | 3. concatenate to a numpy array 32 | 3. to torch Tensor 33 | 4. normalize 34 | """ 35 | def _resize(im, size): 36 | return cv2.resize(im.astype(np.float32)/255., size) 37 | 38 | im_batch = torch.cat([self.norm(_resize(im, self.size)).unsqueeze(0) for im in im_crops], dim=0).float() 39 | return im_batch 40 | 41 | 42 | def __call__(self, im_crops): 43 | im_batch = self._preprocess(im_crops) 44 | with torch.no_grad(): 45 | im_batch = im_batch.to(self.device) 46 | features = self.net(im_batch) 47 | return features.cpu().numpy() 48 | 49 | 50 | if __name__ == '__main__': 51 | img = cv2.imread("demo.jpg")[:,:,(2,1,0)] 52 | extr = Extractor("checkpoint/ckpt.t7") 53 | feature = extr(img) 54 | print(feature.shape) 55 | 56 | -------------------------------------------------------------------------------- /utils/flask_rest_api/README.md: -------------------------------------------------------------------------------- 1 | # Flask REST API 2 | [REST](https://en.wikipedia.org/wiki/Representational_state_transfer) [API](https://en.wikipedia.org/wiki/API)s are commonly used to expose Machine Learning (ML) models to other services. This folder contains an example REST API created using Flask to expose the YOLOv5s model from [PyTorch Hub](https://pytorch.org/hub/ultralytics_yolov5/). 3 | 4 | ## Requirements 5 | 6 | [Flask](https://palletsprojects.com/p/flask/) is required. Install with: 7 | ```shell 8 | $ pip install Flask 9 | ``` 10 | 11 | ## Run 12 | 13 | After Flask installation run: 14 | 15 | ```shell 16 | $ python3 restapi.py --port 5000 17 | ``` 18 | 19 | Then use [curl](https://curl.se/) to perform a request: 20 | 21 | ```shell 22 | $ curl -X POST -F image=@zidane.jpg 'http://localhost:5000/v1/object-detection/yolov5s'` 23 | ``` 24 | 25 | The model inference results are returned as a JSON response: 26 | 27 | ```json 28 | [ 29 | { 30 | "class": 0, 31 | "confidence": 0.8900438547, 32 | "height": 0.9318675399, 33 | "name": "person", 34 | "width": 0.3264600933, 35 | "xcenter": 0.7438579798, 36 | "ycenter": 0.5207948685 37 | }, 38 | { 39 | "class": 0, 40 | "confidence": 0.8440024257, 41 | "height": 0.7155083418, 42 | "name": "person", 43 | "width": 0.6546785235, 44 | "xcenter": 0.427829951, 45 | "ycenter": 0.6334488392 46 | }, 47 | { 48 | "class": 27, 49 | "confidence": 0.3771208823, 50 | "height": 0.3902671337, 51 | "name": "tie", 52 | "width": 0.0696444362, 53 | "xcenter": 0.3675483763, 54 | "ycenter": 0.7991207838 55 | }, 56 | { 57 | "class": 27, 58 | "confidence": 0.3527112305, 59 | "height": 0.1540903747, 60 | "name": "tie", 61 | "width": 0.0336618312, 62 | "xcenter": 0.7814827561, 63 | "ycenter": 0.5065554976 64 | } 65 | ] 66 | ``` 67 | 68 | An example python script to perform inference using [requests](https://docs.python-requests.org/en/master/) is given in `example_request.py` 69 | -------------------------------------------------------------------------------- /models/hub/yolov5-p6.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 3 8 | 9 | # YOLOv5 backbone 10 | backbone: 11 | # [from, number, module, args] 12 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 13 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 14 | [ -1, 3, C3, [ 128 ] ], 15 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 16 | [ -1, 9, C3, [ 256 ] ], 17 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 18 | [ -1, 9, C3, [ 512 ] ], 19 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 20 | [ -1, 3, C3, [ 768 ] ], 21 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 22 | [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ], 23 | [ -1, 3, C3, [ 1024, False ] ], # 11 24 | ] 25 | 26 | # YOLOv5 head 27 | head: 28 | [ [ -1, 1, Conv, [ 768, 1, 1 ] ], 29 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 30 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 31 | [ -1, 3, C3, [ 768, False ] ], # 15 32 | 33 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 34 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 35 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 36 | [ -1, 3, C3, [ 512, False ] ], # 19 37 | 38 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 39 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 40 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 41 | [ -1, 3, C3, [ 256, False ] ], # 23 (P3/8-small) 42 | 43 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 44 | [ [ -1, 20 ], 1, Concat, [ 1 ] ], # cat head P4 45 | [ -1, 3, C3, [ 512, False ] ], # 26 (P4/16-medium) 46 | 47 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 48 | [ [ -1, 16 ], 1, Concat, [ 1 ] ], # cat head P5 49 | [ -1, 3, C3, [ 768, False ] ], # 29 (P5/32-large) 50 | 51 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 52 | [ [ -1, 12 ], 1, Concat, [ 1 ] ], # cat head P6 53 | [ -1, 3, C3, [ 1024, False ] ], # 32 (P5/64-xlarge) 54 | 55 | [ [ 23, 26, 29, 32 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6) 56 | ] 57 | -------------------------------------------------------------------------------- /models/hub/yolov5l6.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [ 19,27, 44,40, 38,94 ] # P3/8 9 | - [ 96,68, 86,152, 180,137 ] # P4/16 10 | - [ 140,301, 303,264, 238,542 ] # P5/32 11 | - [ 436,615, 739,380, 925,792 ] # P6/64 12 | 13 | # YOLOv5 backbone 14 | backbone: 15 | # [from, number, module, args] 16 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 17 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 18 | [ -1, 3, C3, [ 128 ] ], 19 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 20 | [ -1, 9, C3, [ 256 ] ], 21 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 22 | [ -1, 9, C3, [ 512 ] ], 23 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 24 | [ -1, 3, C3, [ 768 ] ], 25 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 26 | [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ], 27 | [ -1, 3, C3, [ 1024, False ] ], # 11 28 | ] 29 | 30 | # YOLOv5 head 31 | head: 32 | [ [ -1, 1, Conv, [ 768, 1, 1 ] ], 33 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 34 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 35 | [ -1, 3, C3, [ 768, False ] ], # 15 36 | 37 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 38 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 39 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 40 | [ -1, 3, C3, [ 512, False ] ], # 19 41 | 42 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 43 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 44 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 45 | [ -1, 3, C3, [ 256, False ] ], # 23 (P3/8-small) 46 | 47 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 48 | [ [ -1, 20 ], 1, Concat, [ 1 ] ], # cat head P4 49 | [ -1, 3, C3, [ 512, False ] ], # 26 (P4/16-medium) 50 | 51 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 52 | [ [ -1, 16 ], 1, Concat, [ 1 ] ], # cat head P5 53 | [ -1, 3, C3, [ 768, False ] ], # 29 (P5/32-large) 54 | 55 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 56 | [ [ -1, 12 ], 1, Concat, [ 1 ] ], # cat head P6 57 | [ -1, 3, C3, [ 1024, False ] ], # 32 (P6/64-xlarge) 58 | 59 | [ [ 23, 26, 29, 32 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6) 60 | ] 61 | -------------------------------------------------------------------------------- /models/hub/yolov5m6.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.67 # model depth multiple 4 | width_multiple: 0.75 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [ 19,27, 44,40, 38,94 ] # P3/8 9 | - [ 96,68, 86,152, 180,137 ] # P4/16 10 | - [ 140,301, 303,264, 238,542 ] # P5/32 11 | - [ 436,615, 739,380, 925,792 ] # P6/64 12 | 13 | # YOLOv5 backbone 14 | backbone: 15 | # [from, number, module, args] 16 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 17 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 18 | [ -1, 3, C3, [ 128 ] ], 19 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 20 | [ -1, 9, C3, [ 256 ] ], 21 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 22 | [ -1, 9, C3, [ 512 ] ], 23 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 24 | [ -1, 3, C3, [ 768 ] ], 25 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 26 | [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ], 27 | [ -1, 3, C3, [ 1024, False ] ], # 11 28 | ] 29 | 30 | # YOLOv5 head 31 | head: 32 | [ [ -1, 1, Conv, [ 768, 1, 1 ] ], 33 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 34 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 35 | [ -1, 3, C3, [ 768, False ] ], # 15 36 | 37 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 38 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 39 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 40 | [ -1, 3, C3, [ 512, False ] ], # 19 41 | 42 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 43 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 44 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 45 | [ -1, 3, C3, [ 256, False ] ], # 23 (P3/8-small) 46 | 47 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 48 | [ [ -1, 20 ], 1, Concat, [ 1 ] ], # cat head P4 49 | [ -1, 3, C3, [ 512, False ] ], # 26 (P4/16-medium) 50 | 51 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 52 | [ [ -1, 16 ], 1, Concat, [ 1 ] ], # cat head P5 53 | [ -1, 3, C3, [ 768, False ] ], # 29 (P5/32-large) 54 | 55 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 56 | [ [ -1, 12 ], 1, Concat, [ 1 ] ], # cat head P6 57 | [ -1, 3, C3, [ 1024, False ] ], # 32 (P6/64-xlarge) 58 | 59 | [ [ 23, 26, 29, 32 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6) 60 | ] 61 | -------------------------------------------------------------------------------- /models/hub/yolov5s6.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.33 # model depth multiple 4 | width_multiple: 0.50 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [ 19,27, 44,40, 38,94 ] # P3/8 9 | - [ 96,68, 86,152, 180,137 ] # P4/16 10 | - [ 140,301, 303,264, 238,542 ] # P5/32 11 | - [ 436,615, 739,380, 925,792 ] # P6/64 12 | 13 | # YOLOv5 backbone 14 | backbone: 15 | # [from, number, module, args] 16 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 17 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 18 | [ -1, 3, C3, [ 128 ] ], 19 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 20 | [ -1, 9, C3, [ 256 ] ], 21 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 22 | [ -1, 9, C3, [ 512 ] ], 23 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 24 | [ -1, 3, C3, [ 768 ] ], 25 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 26 | [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ], 27 | [ -1, 3, C3, [ 1024, False ] ], # 11 28 | ] 29 | 30 | # YOLOv5 head 31 | head: 32 | [ [ -1, 1, Conv, [ 768, 1, 1 ] ], 33 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 34 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 35 | [ -1, 3, C3, [ 768, False ] ], # 15 36 | 37 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 38 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 39 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 40 | [ -1, 3, C3, [ 512, False ] ], # 19 41 | 42 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 43 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 44 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 45 | [ -1, 3, C3, [ 256, False ] ], # 23 (P3/8-small) 46 | 47 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 48 | [ [ -1, 20 ], 1, Concat, [ 1 ] ], # cat head P4 49 | [ -1, 3, C3, [ 512, False ] ], # 26 (P4/16-medium) 50 | 51 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 52 | [ [ -1, 16 ], 1, Concat, [ 1 ] ], # cat head P5 53 | [ -1, 3, C3, [ 768, False ] ], # 29 (P5/32-large) 54 | 55 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 56 | [ [ -1, 12 ], 1, Concat, [ 1 ] ], # cat head P6 57 | [ -1, 3, C3, [ 1024, False ] ], # 32 (P6/64-xlarge) 58 | 59 | [ [ 23, 26, 29, 32 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6) 60 | ] 61 | -------------------------------------------------------------------------------- /models/hub/yolov5x6.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.33 # model depth multiple 4 | width_multiple: 1.25 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [ 19,27, 44,40, 38,94 ] # P3/8 9 | - [ 96,68, 86,152, 180,137 ] # P4/16 10 | - [ 140,301, 303,264, 238,542 ] # P5/32 11 | - [ 436,615, 739,380, 925,792 ] # P6/64 12 | 13 | # YOLOv5 backbone 14 | backbone: 15 | # [from, number, module, args] 16 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 17 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 18 | [ -1, 3, C3, [ 128 ] ], 19 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 20 | [ -1, 9, C3, [ 256 ] ], 21 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 22 | [ -1, 9, C3, [ 512 ] ], 23 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 24 | [ -1, 3, C3, [ 768 ] ], 25 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 26 | [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ], 27 | [ -1, 3, C3, [ 1024, False ] ], # 11 28 | ] 29 | 30 | # YOLOv5 head 31 | head: 32 | [ [ -1, 1, Conv, [ 768, 1, 1 ] ], 33 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 34 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 35 | [ -1, 3, C3, [ 768, False ] ], # 15 36 | 37 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 38 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 39 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 40 | [ -1, 3, C3, [ 512, False ] ], # 19 41 | 42 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 43 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 44 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 45 | [ -1, 3, C3, [ 256, False ] ], # 23 (P3/8-small) 46 | 47 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 48 | [ [ -1, 20 ], 1, Concat, [ 1 ] ], # cat head P4 49 | [ -1, 3, C3, [ 512, False ] ], # 26 (P4/16-medium) 50 | 51 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 52 | [ [ -1, 16 ], 1, Concat, [ 1 ] ], # cat head P5 53 | [ -1, 3, C3, [ 768, False ] ], # 29 (P5/32-large) 54 | 55 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 56 | [ [ -1, 12 ], 1, Concat, [ 1 ] ], # cat head P6 57 | [ -1, 3, C3, [ 1024, False ] ], # 32 (P6/64-xlarge) 58 | 59 | [ [ 23, 26, 29, 32 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6) 60 | ] 61 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/preprocessing.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | import numpy as np 3 | import cv2 4 | 5 | 6 | def non_max_suppression(boxes, max_bbox_overlap, scores=None): 7 | """Suppress overlapping detections. 8 | 9 | Original code from [1]_ has been adapted to include confidence score. 10 | 11 | .. [1] http://www.pyimagesearch.com/2015/02/16/ 12 | faster-non-maximum-suppression-python/ 13 | 14 | Examples 15 | -------- 16 | 17 | >>> boxes = [d.roi for d in detections] 18 | >>> scores = [d.confidence for d in detections] 19 | >>> indices = non_max_suppression(boxes, max_bbox_overlap, scores) 20 | >>> detections = [detections[i] for i in indices] 21 | 22 | Parameters 23 | ---------- 24 | boxes : ndarray 25 | Array of ROIs (x, y, width, height). 26 | max_bbox_overlap : float 27 | ROIs that overlap more than this values are suppressed. 28 | scores : Optional[array_like] 29 | Detector confidence score. 30 | 31 | Returns 32 | ------- 33 | List[int] 34 | Returns indices of detections that have survived non-maxima suppression. 35 | 36 | """ 37 | if len(boxes) == 0: 38 | return [] 39 | 40 | boxes = boxes.astype(np.float) 41 | pick = [] 42 | 43 | x1 = boxes[:, 0] 44 | y1 = boxes[:, 1] 45 | x2 = boxes[:, 2] + boxes[:, 0] 46 | y2 = boxes[:, 3] + boxes[:, 1] 47 | 48 | area = (x2 - x1 + 1) * (y2 - y1 + 1) 49 | if scores is not None: 50 | idxs = np.argsort(scores) 51 | else: 52 | idxs = np.argsort(y2) 53 | 54 | while len(idxs) > 0: 55 | last = len(idxs) - 1 56 | i = idxs[last] 57 | pick.append(i) 58 | 59 | xx1 = np.maximum(x1[i], x1[idxs[:last]]) 60 | yy1 = np.maximum(y1[i], y1[idxs[:last]]) 61 | xx2 = np.minimum(x2[i], x2[idxs[:last]]) 62 | yy2 = np.minimum(y2[i], y2[idxs[:last]]) 63 | 64 | w = np.maximum(0, xx2 - xx1 + 1) 65 | h = np.maximum(0, yy2 - yy1 + 1) 66 | 67 | overlap = (w * h) / area[idxs[:last]] 68 | 69 | idxs = np.delete( 70 | idxs, np.concatenate( 71 | ([last], np.where(overlap > max_bbox_overlap)[0]))) 72 | 73 | return pick 74 | -------------------------------------------------------------------------------- /AIDetector_pytorch.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import numpy as np 3 | from models.experimental import attempt_load 4 | from utils.general import non_max_suppression, scale_coords 5 | from utils.BaseDetector import baseDet 6 | from utils.torch_utils import select_device 7 | from utils.datasets import letterbox 8 | 9 | class Detector(baseDet): 10 | 11 | def __init__(self): 12 | super(Detector, self).__init__() 13 | self.init_model() 14 | self.build_config() 15 | 16 | def init_model(self): 17 | 18 | self.weights = 'weights/yolov5s.pt' 19 | self.device = '0' if torch.cuda.is_available() else 'cpu' 20 | self.device = select_device(self.device) 21 | model = attempt_load(self.weights, map_location=self.device) 22 | model.to(self.device).eval() 23 | model.half() 24 | # torch.save(model, 'test.pt') 25 | self.m = model 26 | self.names = model.module.names if hasattr( 27 | model, 'module') else model.names 28 | 29 | def preprocess(self, img): 30 | 31 | img0 = img.copy() 32 | img = letterbox(img, new_shape=self.img_size)[0] 33 | img = img[:, :, ::-1].transpose(2, 0, 1) 34 | img = np.ascontiguousarray(img) 35 | img = torch.from_numpy(img).to(self.device) 36 | img = img.half() # 半精度 37 | img /= 255.0 # 图像归一化 38 | if img.ndimension() == 3: 39 | img = img.unsqueeze(0) 40 | 41 | return img0, img 42 | 43 | def detect(self, im): 44 | 45 | im0, img = self.preprocess(im) 46 | 47 | pred = self.m(img, augment=False)[0] 48 | pred = pred.float() 49 | pred = non_max_suppression(pred, self.threshold, 0.4) 50 | 51 | pred_boxes = [] 52 | for det in pred: 53 | 54 | if det is not None and len(det): 55 | det[:, :4] = scale_coords( 56 | img.shape[2:], det[:, :4], im0.shape).round() 57 | 58 | for *x, conf, cls_id in det: 59 | lbl = self.names[int(cls_id)] 60 | if not lbl in ['person', 'car', 'truck']: 61 | continue 62 | x1, y1 = int(x[0]), int(x[1]) 63 | x2, y2 = int(x[2]), int(x[3]) 64 | pred_boxes.append( 65 | (x1, y1, x2, y2, lbl, conf)) 66 | 67 | return im, pred_boxes 68 | 69 | -------------------------------------------------------------------------------- /models/hub/yolov5-p7.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 3 8 | 9 | # YOLOv5 backbone 10 | backbone: 11 | # [from, number, module, args] 12 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 13 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 14 | [ -1, 3, C3, [ 128 ] ], 15 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 16 | [ -1, 9, C3, [ 256 ] ], 17 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 18 | [ -1, 9, C3, [ 512 ] ], 19 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 20 | [ -1, 3, C3, [ 768 ] ], 21 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 22 | [ -1, 3, C3, [ 1024 ] ], 23 | [ -1, 1, Conv, [ 1280, 3, 2 ] ], # 11-P7/128 24 | [ -1, 1, SPP, [ 1280, [ 3, 5 ] ] ], 25 | [ -1, 3, C3, [ 1280, False ] ], # 13 26 | ] 27 | 28 | # YOLOv5 head 29 | head: 30 | [ [ -1, 1, Conv, [ 1024, 1, 1 ] ], 31 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 32 | [ [ -1, 10 ], 1, Concat, [ 1 ] ], # cat backbone P6 33 | [ -1, 3, C3, [ 1024, False ] ], # 17 34 | 35 | [ -1, 1, Conv, [ 768, 1, 1 ] ], 36 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 37 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 38 | [ -1, 3, C3, [ 768, False ] ], # 21 39 | 40 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 41 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 42 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 43 | [ -1, 3, C3, [ 512, False ] ], # 25 44 | 45 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 46 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 47 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 48 | [ -1, 3, C3, [ 256, False ] ], # 29 (P3/8-small) 49 | 50 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 51 | [ [ -1, 26 ], 1, Concat, [ 1 ] ], # cat head P4 52 | [ -1, 3, C3, [ 512, False ] ], # 32 (P4/16-medium) 53 | 54 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 55 | [ [ -1, 22 ], 1, Concat, [ 1 ] ], # cat head P5 56 | [ -1, 3, C3, [ 768, False ] ], # 35 (P5/32-large) 57 | 58 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 59 | [ [ -1, 18 ], 1, Concat, [ 1 ] ], # cat head P6 60 | [ -1, 3, C3, [ 1024, False ] ], # 38 (P6/64-xlarge) 61 | 62 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], 63 | [ [ -1, 14 ], 1, Concat, [ 1 ] ], # cat head P7 64 | [ -1, 3, C3, [ 1280, False ] ], # 41 (P7/128-xxlarge) 65 | 66 | [ [ 29, 32, 35, 38, 41 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6, P7) 67 | ] 68 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/test.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.backends.cudnn as cudnn 3 | import torchvision 4 | 5 | import argparse 6 | import os 7 | 8 | from model import Net 9 | 10 | parser = argparse.ArgumentParser(description="Train on market1501") 11 | parser.add_argument("--data-dir",default='data',type=str) 12 | parser.add_argument("--no-cuda",action="store_true") 13 | parser.add_argument("--gpu-id",default=0,type=int) 14 | args = parser.parse_args() 15 | 16 | # device 17 | device = "cuda:{}".format(args.gpu_id) if torch.cuda.is_available() and not args.no_cuda else "cpu" 18 | if torch.cuda.is_available() and not args.no_cuda: 19 | cudnn.benchmark = True 20 | 21 | # data loader 22 | root = args.data_dir 23 | query_dir = os.path.join(root,"query") 24 | gallery_dir = os.path.join(root,"gallery") 25 | transform = torchvision.transforms.Compose([ 26 | torchvision.transforms.Resize((128,64)), 27 | torchvision.transforms.ToTensor(), 28 | torchvision.transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) 29 | ]) 30 | queryloader = torch.utils.data.DataLoader( 31 | torchvision.datasets.ImageFolder(query_dir, transform=transform), 32 | batch_size=64, shuffle=False 33 | ) 34 | galleryloader = torch.utils.data.DataLoader( 35 | torchvision.datasets.ImageFolder(gallery_dir, transform=transform), 36 | batch_size=64, shuffle=False 37 | ) 38 | 39 | # net definition 40 | net = Net(reid=True) 41 | assert os.path.isfile("./checkpoint/ckpt.t7"), "Error: no checkpoint file found!" 42 | print('Loading from checkpoint/ckpt.t7') 43 | checkpoint = torch.load("./checkpoint/ckpt.t7") 44 | net_dict = checkpoint['net_dict'] 45 | net.load_state_dict(net_dict, strict=False) 46 | net.eval() 47 | net.to(device) 48 | 49 | # compute features 50 | query_features = torch.tensor([]).float() 51 | query_labels = torch.tensor([]).long() 52 | gallery_features = torch.tensor([]).float() 53 | gallery_labels = torch.tensor([]).long() 54 | 55 | with torch.no_grad(): 56 | for idx,(inputs,labels) in enumerate(queryloader): 57 | inputs = inputs.to(device) 58 | features = net(inputs).cpu() 59 | query_features = torch.cat((query_features, features), dim=0) 60 | query_labels = torch.cat((query_labels, labels)) 61 | 62 | for idx,(inputs,labels) in enumerate(galleryloader): 63 | inputs = inputs.to(device) 64 | features = net(inputs).cpu() 65 | gallery_features = torch.cat((gallery_features, features), dim=0) 66 | gallery_labels = torch.cat((gallery_labels, labels)) 67 | 68 | gallery_labels -= 2 69 | 70 | # save features 71 | features = { 72 | "qf": query_features, 73 | "ql": query_labels, 74 | "gf": gallery_features, 75 | "gl": gallery_labels 76 | } 77 | torch.save(features,"features.pth") -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/iou_matching.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | from __future__ import absolute_import 3 | import numpy as np 4 | from . import linear_assignment 5 | 6 | 7 | def iou(bbox, candidates): 8 | """Computer intersection over union. 9 | 10 | Parameters 11 | ---------- 12 | bbox : ndarray 13 | A bounding box in format `(top left x, top left y, width, height)`. 14 | candidates : ndarray 15 | A matrix of candidate bounding boxes (one per row) in the same format 16 | as `bbox`. 17 | 18 | Returns 19 | ------- 20 | ndarray 21 | The intersection over union in [0, 1] between the `bbox` and each 22 | candidate. A higher score means a larger fraction of the `bbox` is 23 | occluded by the candidate. 24 | 25 | """ 26 | bbox_tl, bbox_br = bbox[:2], bbox[:2] + bbox[2:] 27 | candidates_tl = candidates[:, :2] 28 | candidates_br = candidates[:, :2] + candidates[:, 2:] 29 | 30 | tl = np.c_[np.maximum(bbox_tl[0], candidates_tl[:, 0])[:, np.newaxis], 31 | np.maximum(bbox_tl[1], candidates_tl[:, 1])[:, np.newaxis]] 32 | br = np.c_[np.minimum(bbox_br[0], candidates_br[:, 0])[:, np.newaxis], 33 | np.minimum(bbox_br[1], candidates_br[:, 1])[:, np.newaxis]] 34 | wh = np.maximum(0., br - tl) 35 | 36 | area_intersection = wh.prod(axis=1) 37 | area_bbox = bbox[2:].prod() 38 | area_candidates = candidates[:, 2:].prod(axis=1) 39 | return area_intersection / (area_bbox + area_candidates - area_intersection) 40 | 41 | 42 | def iou_cost(tracks, detections, track_indices=None, 43 | detection_indices=None): 44 | """An intersection over union distance metric. 45 | 46 | Parameters 47 | ---------- 48 | tracks : List[deep_sort.track.Track] 49 | A list of tracks. 50 | detections : List[deep_sort.detection.Detection] 51 | A list of detections. 52 | track_indices : Optional[List[int]] 53 | A list of indices to tracks that should be matched. Defaults to 54 | all `tracks`. 55 | detection_indices : Optional[List[int]] 56 | A list of indices to detections that should be matched. Defaults 57 | to all `detections`. 58 | 59 | Returns 60 | ------- 61 | ndarray 62 | Returns a cost matrix of shape 63 | len(track_indices), len(detection_indices) where entry (i, j) is 64 | `1 - iou(tracks[track_indices[i]], detections[detection_indices[j]])`. 65 | 66 | """ 67 | if track_indices is None: 68 | track_indices = np.arange(len(tracks)) 69 | if detection_indices is None: 70 | detection_indices = np.arange(len(detections)) 71 | 72 | cost_matrix = np.zeros((len(track_indices), len(detection_indices))) 73 | for row, track_idx in enumerate(track_indices): 74 | if tracks[track_idx].time_since_update > 1: 75 | cost_matrix[row, :] = linear_assignment.INFTY_COST 76 | continue 77 | 78 | bbox = tracks[track_idx].to_tlwh() 79 | candidates = np.asarray([detections[i].tlwh for i in detection_indices]) 80 | cost_matrix[row, :] = 1. - iou(bbox, candidates) 81 | return cost_matrix 82 | -------------------------------------------------------------------------------- /tracker.py: -------------------------------------------------------------------------------- 1 | from deep_sort.utils.parser import get_config 2 | from deep_sort.deep_sort import DeepSort 3 | import torch 4 | import cv2 5 | 6 | palette = (2 ** 11 - 1, 2 ** 15 - 1, 2 ** 20 - 1) 7 | cfg = get_config() 8 | cfg.merge_from_file("deep_sort/configs/deep_sort.yaml") 9 | deepsort = DeepSort(cfg.DEEPSORT.REID_CKPT, 10 | max_dist=cfg.DEEPSORT.MAX_DIST, min_confidence=cfg.DEEPSORT.MIN_CONFIDENCE, 11 | nms_max_overlap=cfg.DEEPSORT.NMS_MAX_OVERLAP, max_iou_distance=cfg.DEEPSORT.MAX_IOU_DISTANCE, 12 | max_age=cfg.DEEPSORT.MAX_AGE, n_init=cfg.DEEPSORT.N_INIT, nn_budget=cfg.DEEPSORT.NN_BUDGET, 13 | use_cuda=True) 14 | 15 | 16 | def plot_bboxes(image, bboxes, line_thickness=None): 17 | # Plots one bounding box on image img 18 | tl = line_thickness or round( 19 | 0.002 * (image.shape[0] + image.shape[1]) / 2) + 1 # line/font thickness 20 | for (x1, y1, x2, y2, cls_id, pos_id) in bboxes: 21 | if cls_id in ['person']: 22 | color = (0, 0, 255) 23 | else: 24 | color = (0, 255, 0) 25 | c1, c2 = (x1, y1), (x2, y2) 26 | cv2.rectangle(image, c1, c2, color, thickness=tl, lineType=cv2.LINE_AA) 27 | tf = max(tl - 1, 1) # font thickness 28 | t_size = cv2.getTextSize(cls_id, 0, fontScale=tl / 3, thickness=tf)[0] 29 | c2 = c1[0] + t_size[0], c1[1] - t_size[1] - 3 30 | cv2.rectangle(image, c1, c2, color, -1, cv2.LINE_AA) # filled 31 | cv2.putText(image, '{} ID-{}'.format(cls_id, pos_id), (c1[0], c1[1] - 2), 0, tl / 3, 32 | [225, 255, 255], thickness=tf, lineType=cv2.LINE_AA) 33 | 34 | return image 35 | 36 | 37 | def update_tracker(target_detector, image): 38 | 39 | new_faces = [] 40 | _, bboxes = target_detector.detect(image) 41 | 42 | bbox_xywh = [] 43 | confs = [] 44 | clss = [] 45 | 46 | for x1, y1, x2, y2, cls_id, conf in bboxes: 47 | 48 | obj = [ 49 | int((x1+x2)/2), int((y1+y2)/2), 50 | x2-x1, y2-y1 51 | ] 52 | bbox_xywh.append(obj) 53 | confs.append(conf) 54 | clss.append(cls_id) 55 | 56 | xywhs = torch.Tensor(bbox_xywh) 57 | confss = torch.Tensor(confs) 58 | 59 | outputs = deepsort.update(xywhs, confss, clss, image) 60 | 61 | bboxes2draw = [] 62 | face_bboxes = [] 63 | current_ids = [] 64 | for value in list(outputs): 65 | x1, y1, x2, y2, cls_, track_id = value 66 | bboxes2draw.append( 67 | (x1, y1, x2, y2, cls_, track_id) 68 | ) 69 | current_ids.append(track_id) 70 | if cls_ == 'face': 71 | if not track_id in target_detector.faceTracker: 72 | target_detector.faceTracker[track_id] = 0 73 | face = image[y1:y2, x1:x2] 74 | new_faces.append((face, track_id)) 75 | face_bboxes.append( 76 | (x1, y1, x2, y2) 77 | ) 78 | 79 | ids2delete = [] 80 | for history_id in target_detector.faceTracker: 81 | if not history_id in current_ids: 82 | target_detector.faceTracker[history_id] -= 1 83 | if target_detector.faceTracker[history_id] < -5: 84 | ids2delete.append(history_id) 85 | 86 | for ids in ids2delete: 87 | target_detector.faceTracker.pop(ids) 88 | print('-[INFO] Delete track id:', ids) 89 | 90 | image = plot_bboxes(image, bboxes2draw) 91 | 92 | return image, new_faces, face_bboxes 93 | -------------------------------------------------------------------------------- /models/hub/anchors.yaml: -------------------------------------------------------------------------------- 1 | # Default YOLOv5 anchors for COCO data 2 | 3 | 4 | # P5 ------------------------------------------------------------------------------------------------------------------- 5 | # P5-640: 6 | anchors_p5_640: 7 | - [ 10,13, 16,30, 33,23 ] # P3/8 8 | - [ 30,61, 62,45, 59,119 ] # P4/16 9 | - [ 116,90, 156,198, 373,326 ] # P5/32 10 | 11 | 12 | # P6 ------------------------------------------------------------------------------------------------------------------- 13 | # P6-640: thr=0.25: 0.9964 BPR, 5.54 anchors past thr, n=12, img_size=640, metric_all=0.281/0.716-mean/best, past_thr=0.469-mean: 9,11, 21,19, 17,41, 43,32, 39,70, 86,64, 65,131, 134,130, 120,265, 282,180, 247,354, 512,387 14 | anchors_p6_640: 15 | - [ 9,11, 21,19, 17,41 ] # P3/8 16 | - [ 43,32, 39,70, 86,64 ] # P4/16 17 | - [ 65,131, 134,130, 120,265 ] # P5/32 18 | - [ 282,180, 247,354, 512,387 ] # P6/64 19 | 20 | # P6-1280: thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1280, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 19,27, 44,40, 38,94, 96,68, 86,152, 180,137, 140,301, 303,264, 238,542, 436,615, 739,380, 925,792 21 | anchors_p6_1280: 22 | - [ 19,27, 44,40, 38,94 ] # P3/8 23 | - [ 96,68, 86,152, 180,137 ] # P4/16 24 | - [ 140,301, 303,264, 238,542 ] # P5/32 25 | - [ 436,615, 739,380, 925,792 ] # P6/64 26 | 27 | # P6-1920: thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1920, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 28,41, 67,59, 57,141, 144,103, 129,227, 270,205, 209,452, 455,396, 358,812, 653,922, 1109,570, 1387,1187 28 | anchors_p6_1920: 29 | - [ 28,41, 67,59, 57,141 ] # P3/8 30 | - [ 144,103, 129,227, 270,205 ] # P4/16 31 | - [ 209,452, 455,396, 358,812 ] # P5/32 32 | - [ 653,922, 1109,570, 1387,1187 ] # P6/64 33 | 34 | 35 | # P7 ------------------------------------------------------------------------------------------------------------------- 36 | # P7-640: thr=0.25: 0.9962 BPR, 6.76 anchors past thr, n=15, img_size=640, metric_all=0.275/0.733-mean/best, past_thr=0.466-mean: 11,11, 13,30, 29,20, 30,46, 61,38, 39,92, 78,80, 146,66, 79,163, 149,150, 321,143, 157,303, 257,402, 359,290, 524,372 37 | anchors_p7_640: 38 | - [ 11,11, 13,30, 29,20 ] # P3/8 39 | - [ 30,46, 61,38, 39,92 ] # P4/16 40 | - [ 78,80, 146,66, 79,163 ] # P5/32 41 | - [ 149,150, 321,143, 157,303 ] # P6/64 42 | - [ 257,402, 359,290, 524,372 ] # P7/128 43 | 44 | # P7-1280: thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1280, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 19,22, 54,36, 32,77, 70,83, 138,71, 75,173, 165,159, 148,334, 375,151, 334,317, 251,626, 499,474, 750,326, 534,814, 1079,818 45 | anchors_p7_1280: 46 | - [ 19,22, 54,36, 32,77 ] # P3/8 47 | - [ 70,83, 138,71, 75,173 ] # P4/16 48 | - [ 165,159, 148,334, 375,151 ] # P5/32 49 | - [ 334,317, 251,626, 499,474 ] # P6/64 50 | - [ 750,326, 534,814, 1079,818 ] # P7/128 51 | 52 | # P7-1920: thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1920, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 29,34, 81,55, 47,115, 105,124, 207,107, 113,259, 247,238, 222,500, 563,227, 501,476, 376,939, 749,711, 1126,489, 801,1222, 1618,1227 53 | anchors_p7_1920: 54 | - [ 29,34, 81,55, 47,115 ] # P3/8 55 | - [ 105,124, 207,107, 113,259 ] # P4/16 56 | - [ 247,238, 222,500, 563,227 ] # P5/32 57 | - [ 501,476, 376,939, 749,711 ] # P6/64 58 | - [ 1126,489, 801,1222, 1618,1227 ] # P7/128 59 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/model.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.nn as nn 3 | import torch.nn.functional as F 4 | 5 | class BasicBlock(nn.Module): 6 | def __init__(self, c_in, c_out,is_downsample=False): 7 | super(BasicBlock,self).__init__() 8 | self.is_downsample = is_downsample 9 | if is_downsample: 10 | self.conv1 = nn.Conv2d(c_in, c_out, 3, stride=2, padding=1, bias=False) 11 | else: 12 | self.conv1 = nn.Conv2d(c_in, c_out, 3, stride=1, padding=1, bias=False) 13 | self.bn1 = nn.BatchNorm2d(c_out) 14 | self.relu = nn.ReLU(True) 15 | self.conv2 = nn.Conv2d(c_out,c_out,3,stride=1,padding=1, bias=False) 16 | self.bn2 = nn.BatchNorm2d(c_out) 17 | if is_downsample: 18 | self.downsample = nn.Sequential( 19 | nn.Conv2d(c_in, c_out, 1, stride=2, bias=False), 20 | nn.BatchNorm2d(c_out) 21 | ) 22 | elif c_in != c_out: 23 | self.downsample = nn.Sequential( 24 | nn.Conv2d(c_in, c_out, 1, stride=1, bias=False), 25 | nn.BatchNorm2d(c_out) 26 | ) 27 | self.is_downsample = True 28 | 29 | def forward(self,x): 30 | y = self.conv1(x) 31 | y = self.bn1(y) 32 | y = self.relu(y) 33 | y = self.conv2(y) 34 | y = self.bn2(y) 35 | if self.is_downsample: 36 | x = self.downsample(x) 37 | return F.relu(x.add(y),True) 38 | 39 | def make_layers(c_in,c_out,repeat_times, is_downsample=False): 40 | blocks = [] 41 | for i in range(repeat_times): 42 | if i ==0: 43 | blocks += [BasicBlock(c_in,c_out, is_downsample=is_downsample),] 44 | else: 45 | blocks += [BasicBlock(c_out,c_out),] 46 | return nn.Sequential(*blocks) 47 | 48 | class Net(nn.Module): 49 | def __init__(self, num_classes=751 ,reid=False): 50 | super(Net,self).__init__() 51 | # 3 128 64 52 | self.conv = nn.Sequential( 53 | nn.Conv2d(3,64,3,stride=1,padding=1), 54 | nn.BatchNorm2d(64), 55 | nn.ReLU(inplace=True), 56 | # nn.Conv2d(32,32,3,stride=1,padding=1), 57 | # nn.BatchNorm2d(32), 58 | # nn.ReLU(inplace=True), 59 | nn.MaxPool2d(3,2,padding=1), 60 | ) 61 | # 32 64 32 62 | self.layer1 = make_layers(64,64,2,False) 63 | # 32 64 32 64 | self.layer2 = make_layers(64,128,2,True) 65 | # 64 32 16 66 | self.layer3 = make_layers(128,256,2,True) 67 | # 128 16 8 68 | self.layer4 = make_layers(256,512,2,True) 69 | # 256 8 4 70 | self.avgpool = nn.AvgPool2d((8,4),1) 71 | # 256 1 1 72 | self.reid = reid 73 | self.classifier = nn.Sequential( 74 | nn.Linear(512, 256), 75 | nn.BatchNorm1d(256), 76 | nn.ReLU(inplace=True), 77 | nn.Dropout(), 78 | nn.Linear(256, num_classes), 79 | ) 80 | 81 | def forward(self, x): 82 | x = self.conv(x) 83 | x = self.layer1(x) 84 | x = self.layer2(x) 85 | x = self.layer3(x) 86 | x = self.layer4(x) 87 | x = self.avgpool(x) 88 | x = x.view(x.size(0),-1) 89 | # B x 128 90 | if self.reid: 91 | x = x.div(x.norm(p=2,dim=1,keepdim=True)) 92 | return x 93 | # classifier 94 | x = self.classifier(x) 95 | return x 96 | 97 | 98 | if __name__ == '__main__': 99 | net = Net() 100 | x = torch.randn(4,3,128,64) 101 | y = net(x) 102 | import ipdb; ipdb.set_trace() 103 | 104 | 105 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/original_model.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.nn as nn 3 | import torch.nn.functional as F 4 | 5 | class BasicBlock(nn.Module): 6 | def __init__(self, c_in, c_out,is_downsample=False): 7 | super(BasicBlock,self).__init__() 8 | self.is_downsample = is_downsample 9 | if is_downsample: 10 | self.conv1 = nn.Conv2d(c_in, c_out, 3, stride=2, padding=1, bias=False) 11 | else: 12 | self.conv1 = nn.Conv2d(c_in, c_out, 3, stride=1, padding=1, bias=False) 13 | self.bn1 = nn.BatchNorm2d(c_out) 14 | self.relu = nn.ReLU(True) 15 | self.conv2 = nn.Conv2d(c_out,c_out,3,stride=1,padding=1, bias=False) 16 | self.bn2 = nn.BatchNorm2d(c_out) 17 | if is_downsample: 18 | self.downsample = nn.Sequential( 19 | nn.Conv2d(c_in, c_out, 1, stride=2, bias=False), 20 | nn.BatchNorm2d(c_out) 21 | ) 22 | elif c_in != c_out: 23 | self.downsample = nn.Sequential( 24 | nn.Conv2d(c_in, c_out, 1, stride=1, bias=False), 25 | nn.BatchNorm2d(c_out) 26 | ) 27 | self.is_downsample = True 28 | 29 | def forward(self,x): 30 | y = self.conv1(x) 31 | y = self.bn1(y) 32 | y = self.relu(y) 33 | y = self.conv2(y) 34 | y = self.bn2(y) 35 | if self.is_downsample: 36 | x = self.downsample(x) 37 | return F.relu(x.add(y),True) 38 | 39 | def make_layers(c_in,c_out,repeat_times, is_downsample=False): 40 | blocks = [] 41 | for i in range(repeat_times): 42 | if i ==0: 43 | blocks += [BasicBlock(c_in,c_out, is_downsample=is_downsample),] 44 | else: 45 | blocks += [BasicBlock(c_out,c_out),] 46 | return nn.Sequential(*blocks) 47 | 48 | class Net(nn.Module): 49 | def __init__(self, num_classes=625 ,reid=False): 50 | super(Net,self).__init__() 51 | # 3 128 64 52 | self.conv = nn.Sequential( 53 | nn.Conv2d(3,32,3,stride=1,padding=1), 54 | nn.BatchNorm2d(32), 55 | nn.ELU(inplace=True), 56 | nn.Conv2d(32,32,3,stride=1,padding=1), 57 | nn.BatchNorm2d(32), 58 | nn.ELU(inplace=True), 59 | nn.MaxPool2d(3,2,padding=1), 60 | ) 61 | # 32 64 32 62 | self.layer1 = make_layers(32,32,2,False) 63 | # 32 64 32 64 | self.layer2 = make_layers(32,64,2,True) 65 | # 64 32 16 66 | self.layer3 = make_layers(64,128,2,True) 67 | # 128 16 8 68 | self.dense = nn.Sequential( 69 | nn.Dropout(p=0.6), 70 | nn.Linear(128*16*8, 128), 71 | nn.BatchNorm1d(128), 72 | nn.ELU(inplace=True) 73 | ) 74 | # 256 1 1 75 | self.reid = reid 76 | self.batch_norm = nn.BatchNorm1d(128) 77 | self.classifier = nn.Sequential( 78 | nn.Linear(128, num_classes), 79 | ) 80 | 81 | def forward(self, x): 82 | x = self.conv(x) 83 | x = self.layer1(x) 84 | x = self.layer2(x) 85 | x = self.layer3(x) 86 | 87 | x = x.view(x.size(0),-1) 88 | if self.reid: 89 | x = self.dense[0](x) 90 | x = self.dense[1](x) 91 | x = x.div(x.norm(p=2,dim=1,keepdim=True)) 92 | return x 93 | x = self.dense(x) 94 | # B x 128 95 | # classifier 96 | x = self.classifier(x) 97 | return x 98 | 99 | 100 | if __name__ == '__main__': 101 | net = Net(reid=True) 102 | x = torch.randn(4,3,128,64) 103 | y = net(x) 104 | import ipdb; ipdb.set_trace() 105 | 106 | 107 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep_sort.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import torch 3 | 4 | from .deep.feature_extractor import Extractor 5 | from .sort.nn_matching import NearestNeighborDistanceMetric 6 | from .sort.preprocessing import non_max_suppression 7 | from .sort.detection import Detection 8 | from .sort.tracker import Tracker 9 | 10 | 11 | __all__ = ['DeepSort'] 12 | 13 | 14 | class DeepSort(object): 15 | def __init__(self, model_path, max_dist=0.2, min_confidence=0.3, nms_max_overlap=1.0, max_iou_distance=0.7, max_age=70, n_init=3, nn_budget=100, use_cuda=True): 16 | self.min_confidence = min_confidence 17 | self.nms_max_overlap = nms_max_overlap 18 | 19 | self.extractor = Extractor(model_path, use_cuda=use_cuda) 20 | 21 | max_cosine_distance = max_dist 22 | nn_budget = 100 23 | metric = NearestNeighborDistanceMetric( 24 | "cosine", max_cosine_distance, nn_budget) 25 | self.tracker = Tracker( 26 | metric, max_iou_distance=max_iou_distance, max_age=max_age, n_init=n_init) 27 | 28 | def update(self, bbox_xywh, confidences, clss, ori_img): 29 | self.height, self.width = ori_img.shape[:2] 30 | # generate detections 31 | features = self._get_features(bbox_xywh, ori_img) 32 | bbox_tlwh = self._xywh_to_tlwh(bbox_xywh) 33 | detections = [Detection(bbox_tlwh[i], clss[i], conf, features[i]) for i, conf in enumerate( 34 | confidences) if conf > self.min_confidence] 35 | # update tracker 36 | self.tracker.predict() 37 | self.tracker.update(detections) 38 | 39 | # output bbox identities 40 | outputs = [] 41 | for track in self.tracker.tracks: 42 | if not track.is_confirmed() or track.time_since_update > 1: 43 | continue 44 | box = track.to_tlwh() 45 | x1, y1, x2, y2 = self._tlwh_to_xyxy(box) 46 | outputs.append((x1, y1, x2, y2, track.cls_, track.track_id)) 47 | return outputs 48 | 49 | @staticmethod 50 | def _xywh_to_tlwh(bbox_xywh): 51 | if isinstance(bbox_xywh, np.ndarray): 52 | bbox_tlwh = bbox_xywh.copy() 53 | elif isinstance(bbox_xywh, torch.Tensor): 54 | bbox_tlwh = bbox_xywh.clone() 55 | bbox_tlwh[:, 0] = bbox_xywh[:, 0] - bbox_xywh[:, 2]/2. 56 | bbox_tlwh[:, 1] = bbox_xywh[:, 1] - bbox_xywh[:, 3]/2. 57 | return bbox_tlwh 58 | 59 | def _xywh_to_xyxy(self, bbox_xywh): 60 | x, y, w, h = bbox_xywh 61 | x1 = max(int(x-w/2), 0) 62 | x2 = min(int(x+w/2), self.width-1) 63 | y1 = max(int(y-h/2), 0) 64 | y2 = min(int(y+h/2), self.height-1) 65 | return x1, y1, x2, y2 66 | 67 | def _tlwh_to_xyxy(self, bbox_tlwh): 68 | """ 69 | TODO: 70 | Convert bbox from xtl_ytl_w_h to xc_yc_w_h 71 | Thanks JieChen91@github.com for reporting this bug! 72 | """ 73 | x, y, w, h = bbox_tlwh 74 | x1 = max(int(x), 0) 75 | x2 = min(int(x+w), self.width-1) 76 | y1 = max(int(y), 0) 77 | y2 = min(int(y+h), self.height-1) 78 | return x1, y1, x2, y2 79 | 80 | def _xyxy_to_tlwh(self, bbox_xyxy): 81 | x1, y1, x2, y2 = bbox_xyxy 82 | 83 | t = x1 84 | l = y1 85 | w = int(x2-x1) 86 | h = int(y2-y1) 87 | return t, l, w, h 88 | 89 | def _get_features(self, bbox_xywh, ori_img): 90 | im_crops = [] 91 | for box in bbox_xywh: 92 | x1, y1, x2, y2 = self._xywh_to_xyxy(box) 93 | im = ori_img[y1:y2, x1:x2] 94 | im_crops.append(im) 95 | if im_crops: 96 | features = self.extractor(im_crops) 97 | else: 98 | features = np.array([]) 99 | return features 100 | -------------------------------------------------------------------------------- /deep_sort/utils/evaluation.py: -------------------------------------------------------------------------------- 1 | import os 2 | import numpy as np 3 | import copy 4 | import motmetrics as mm 5 | mm.lap.default_solver = 'lap' 6 | from utils.io import read_results, unzip_objs 7 | 8 | 9 | class Evaluator(object): 10 | 11 | def __init__(self, data_root, seq_name, data_type): 12 | self.data_root = data_root 13 | self.seq_name = seq_name 14 | self.data_type = data_type 15 | 16 | self.load_annotations() 17 | self.reset_accumulator() 18 | 19 | def load_annotations(self): 20 | assert self.data_type == 'mot' 21 | 22 | gt_filename = os.path.join(self.data_root, self.seq_name, 'gt', 'gt.txt') 23 | self.gt_frame_dict = read_results(gt_filename, self.data_type, is_gt=True) 24 | self.gt_ignore_frame_dict = read_results(gt_filename, self.data_type, is_ignore=True) 25 | 26 | def reset_accumulator(self): 27 | self.acc = mm.MOTAccumulator(auto_id=True) 28 | 29 | def eval_frame(self, frame_id, trk_tlwhs, trk_ids, rtn_events=False): 30 | # results 31 | trk_tlwhs = np.copy(trk_tlwhs) 32 | trk_ids = np.copy(trk_ids) 33 | 34 | # gts 35 | gt_objs = self.gt_frame_dict.get(frame_id, []) 36 | gt_tlwhs, gt_ids = unzip_objs(gt_objs)[:2] 37 | 38 | # ignore boxes 39 | ignore_objs = self.gt_ignore_frame_dict.get(frame_id, []) 40 | ignore_tlwhs = unzip_objs(ignore_objs)[0] 41 | 42 | 43 | # remove ignored results 44 | keep = np.ones(len(trk_tlwhs), dtype=bool) 45 | iou_distance = mm.distances.iou_matrix(ignore_tlwhs, trk_tlwhs, max_iou=0.5) 46 | if len(iou_distance) > 0: 47 | match_is, match_js = mm.lap.linear_sum_assignment(iou_distance) 48 | match_is, match_js = map(lambda a: np.asarray(a, dtype=int), [match_is, match_js]) 49 | match_ious = iou_distance[match_is, match_js] 50 | 51 | match_js = np.asarray(match_js, dtype=int) 52 | match_js = match_js[np.logical_not(np.isnan(match_ious))] 53 | keep[match_js] = False 54 | trk_tlwhs = trk_tlwhs[keep] 55 | trk_ids = trk_ids[keep] 56 | 57 | # get distance matrix 58 | iou_distance = mm.distances.iou_matrix(gt_tlwhs, trk_tlwhs, max_iou=0.5) 59 | 60 | # acc 61 | self.acc.update(gt_ids, trk_ids, iou_distance) 62 | 63 | if rtn_events and iou_distance.size > 0 and hasattr(self.acc, 'last_mot_events'): 64 | events = self.acc.last_mot_events # only supported by https://github.com/longcw/py-motmetrics 65 | else: 66 | events = None 67 | return events 68 | 69 | def eval_file(self, filename): 70 | self.reset_accumulator() 71 | 72 | result_frame_dict = read_results(filename, self.data_type, is_gt=False) 73 | frames = sorted(list(set(self.gt_frame_dict.keys()) | set(result_frame_dict.keys()))) 74 | for frame_id in frames: 75 | trk_objs = result_frame_dict.get(frame_id, []) 76 | trk_tlwhs, trk_ids = unzip_objs(trk_objs)[:2] 77 | self.eval_frame(frame_id, trk_tlwhs, trk_ids, rtn_events=False) 78 | 79 | return self.acc 80 | 81 | @staticmethod 82 | def get_summary(accs, names, metrics=('mota', 'num_switches', 'idp', 'idr', 'idf1', 'precision', 'recall')): 83 | names = copy.deepcopy(names) 84 | if metrics is None: 85 | metrics = mm.metrics.motchallenge_metrics 86 | metrics = copy.deepcopy(metrics) 87 | 88 | mh = mm.metrics.create() 89 | summary = mh.compute_many( 90 | accs, 91 | metrics=metrics, 92 | names=names, 93 | generate_overall=True 94 | ) 95 | 96 | return summary 97 | 98 | @staticmethod 99 | def save_summary(summary, filename): 100 | import pandas as pd 101 | writer = pd.ExcelWriter(filename) 102 | summary.to_excel(writer) 103 | writer.save() 104 | -------------------------------------------------------------------------------- /utils/activations.py: -------------------------------------------------------------------------------- 1 | # Activation functions 2 | 3 | import torch 4 | import torch.nn as nn 5 | import torch.nn.functional as F 6 | 7 | 8 | # SiLU https://arxiv.org/pdf/1606.08415.pdf ---------------------------------------------------------------------------- 9 | class SiLU(nn.Module): # export-friendly version of nn.SiLU() 10 | @staticmethod 11 | def forward(x): 12 | return x * torch.sigmoid(x) 13 | 14 | 15 | class Hardswish(nn.Module): # export-friendly version of nn.Hardswish() 16 | @staticmethod 17 | def forward(x): 18 | # return x * F.hardsigmoid(x) # for torchscript and CoreML 19 | return x * F.hardtanh(x + 3, 0., 6.) / 6. # for torchscript, CoreML and ONNX 20 | 21 | 22 | # Mish https://github.com/digantamisra98/Mish -------------------------------------------------------------------------- 23 | class Mish(nn.Module): 24 | @staticmethod 25 | def forward(x): 26 | return x * F.softplus(x).tanh() 27 | 28 | 29 | class MemoryEfficientMish(nn.Module): 30 | class F(torch.autograd.Function): 31 | @staticmethod 32 | def forward(ctx, x): 33 | ctx.save_for_backward(x) 34 | return x.mul(torch.tanh(F.softplus(x))) # x * tanh(ln(1 + exp(x))) 35 | 36 | @staticmethod 37 | def backward(ctx, grad_output): 38 | x = ctx.saved_tensors[0] 39 | sx = torch.sigmoid(x) 40 | fx = F.softplus(x).tanh() 41 | return grad_output * (fx + x * sx * (1 - fx * fx)) 42 | 43 | def forward(self, x): 44 | return self.F.apply(x) 45 | 46 | 47 | # FReLU https://arxiv.org/abs/2007.11824 ------------------------------------------------------------------------------- 48 | class FReLU(nn.Module): 49 | def __init__(self, c1, k=3): # ch_in, kernel 50 | super().__init__() 51 | self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False) 52 | self.bn = nn.BatchNorm2d(c1) 53 | 54 | def forward(self, x): 55 | return torch.max(x, self.bn(self.conv(x))) 56 | 57 | 58 | # ACON https://arxiv.org/pdf/2009.04759.pdf ---------------------------------------------------------------------------- 59 | class AconC(nn.Module): 60 | r""" ACON activation (activate or not). 61 | AconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is a learnable parameter 62 | according to "Activate or Not: Learning Customized Activation" . 63 | """ 64 | 65 | def __init__(self, c1): 66 | super().__init__() 67 | self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1)) 68 | self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1)) 69 | self.beta = nn.Parameter(torch.ones(1, c1, 1, 1)) 70 | 71 | def forward(self, x): 72 | dpx = (self.p1 - self.p2) * x 73 | return dpx * torch.sigmoid(self.beta * dpx) + self.p2 * x 74 | 75 | 76 | class MetaAconC(nn.Module): 77 | r""" ACON activation (activate or not). 78 | MetaAconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is generated by a small network 79 | according to "Activate or Not: Learning Customized Activation" . 80 | """ 81 | 82 | def __init__(self, c1, k=1, s=1, r=16): # ch_in, kernel, stride, r 83 | super().__init__() 84 | c2 = max(r, c1 // r) 85 | self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1)) 86 | self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1)) 87 | self.fc1 = nn.Conv2d(c1, c2, k, s, bias=True) 88 | self.fc2 = nn.Conv2d(c2, c1, k, s, bias=True) 89 | # self.bn1 = nn.BatchNorm2d(c2) 90 | # self.bn2 = nn.BatchNorm2d(c1) 91 | 92 | def forward(self, x): 93 | y = x.mean(dim=2, keepdims=True).mean(dim=3, keepdims=True) 94 | # batch-size 1 bug/instabilities https://github.com/ultralytics/yolov5/issues/2891 95 | # beta = torch.sigmoid(self.bn2(self.fc2(self.bn1(self.fc1(y))))) # bug/unstable 96 | beta = torch.sigmoid(self.fc2(self.fc1(y))) # bug patch BN layers removed 97 | dpx = (self.p1 - self.p2) * x 98 | return dpx * torch.sigmoid(beta * dpx) + self.p2 * x 99 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # 本文禁止转载! 2 | 3 | 4 | 本文地址：[https://blog.csdn.net/WhiffeYF/article/details/121346483](https://blog.csdn.net/WhiffeYF/article/details/121346483) 5 | 6 | # 项目简介： 7 | 使用YOLOv5+Deepsort实现车辆行人追踪和计数，代码封装成一个Detector类，更容易嵌入到自己的项目中。 8 | 9 | 代码地址（欢迎star）： 10 | 11 | [https://github.com/Sharpiless/yolov5-deepsort/](https://github.com/Sharpiless/yolov5-deepsort/) 12 | 13 | 最终效果： 14 | ![在这里插入图片描述](https://github.com/Whiffe/yolov5-deepsort/blob/main/image.jpg) 15 | # YOLOv5检测器： 16 | 17 | ```python 18 | class Detector(baseDet): 19 | 20 | def __init__(self): 21 | super(Detector, self).__init__() 22 | self.init_model() 23 | self.build_config() 24 | 25 | def init_model(self): 26 | 27 | self.weights = 'weights/yolov5m.pt' 28 | self.device = '0' if torch.cuda.is_available() else 'cpu' 29 | self.device = select_device(self.device) 30 | model = attempt_load(self.weights, map_location=self.device) 31 | model.to(self.device).eval() 32 | model.half() 33 | # torch.save(model, 'test.pt') 34 | self.m = model 35 | self.names = model.module.names if hasattr( 36 | model, 'module') else model.names 37 | 38 | def preprocess(self, img): 39 | 40 | img0 = img.copy() 41 | img = letterbox(img, new_shape=self.img_size)[0] 42 | img = img[:, :, ::-1].transpose(2, 0, 1) 43 | img = np.ascontiguousarray(img) 44 | img = torch.from_numpy(img).to(self.device) 45 | img = img.half() # 半精度 46 | img /= 255.0 # 图像归一化 47 | if img.ndimension() == 3: 48 | img = img.unsqueeze(0) 49 | 50 | return img0, img 51 | 52 | def detect(self, im): 53 | 54 | im0, img = self.preprocess(im) 55 | 56 | pred = self.m(img, augment=False)[0] 57 | pred = pred.float() 58 | pred = non_max_suppression(pred, self.threshold, 0.4) 59 | 60 | pred_boxes = [] 61 | for det in pred: 62 | 63 | if det is not None and len(det): 64 | det[:, :4] = scale_coords( 65 | img.shape[2:], det[:, :4], im0.shape).round() 66 | 67 | for *x, conf, cls_id in det: 68 | lbl = self.names[int(cls_id)] 69 | if not lbl in ['person', 'car', 'truck']: 70 | continue 71 | x1, y1 = int(x[0]), int(x[1]) 72 | x2, y2 = int(x[2]), int(x[3]) 73 | pred_boxes.append( 74 | (x1, y1, x2, y2, lbl, conf)) 75 | 76 | return im, pred_boxes 77 | 78 | ``` 79 | 80 | 调用 self.detect 方法返回图像和预测结果 81 | 82 | # DeepSort追踪器： 83 | 84 | ```python 85 | deepsort = DeepSort(cfg.DEEPSORT.REID_CKPT, 86 | max_dist=cfg.DEEPSORT.MAX_DIST, min_confidence=cfg.DEEPSORT.MIN_CONFIDENCE, 87 | nms_max_overlap=cfg.DEEPSORT.NMS_MAX_OVERLAP, max_iou_distance=cfg.DEEPSORT.MAX_IOU_DISTANCE, 88 | max_age=cfg.DEEPSORT.MAX_AGE, n_init=cfg.DEEPSORT.N_INIT, nn_budget=cfg.DEEPSORT.NN_BUDGET, 89 | use_cuda=True) 90 | ``` 91 | 92 | 调用 self.update 方法更新追踪结果 93 | 94 | # 运行demo： 95 | 96 | ```bash 97 | python demo.py 98 | ``` 99 | 100 | # 训练自己的模型： 101 | 参考我的另一篇博客： 102 | 103 | [【小白CV】手把手教你用YOLOv5训练自己的数据集（从Windows环境配置到模型部署）](https://blog.csdn.net/weixin_44936889/article/details/110661862) 104 | 105 | 训练好后放到 weights 文件夹下 106 | 107 | # 调用接口： 108 | 109 | ## 创建检测器： 110 | 111 | ```python 112 | from AIDetector_pytorch import Detector 113 | 114 | det = Detector() 115 | ``` 116 | 117 | ## 调用检测接口： 118 | 119 | ```python 120 | result = det.feedCap(im) 121 | ``` 122 | 123 | 其中 im 为 BGR 图像 124 | 125 | 返回的 result 是字典，result['frame'] 返回可视化后的图像 126 | 127 | # 联系作者： 128 | 129 | > B站：[https://space.bilibili.com/470550823](https://www.bilibili.com/video/BV1A44y1v7ig/) 130 | 131 | > CSDN：[https://blog.csdn.net/weixin_44936889](https://blog.csdn.net/WhiffeYF/article/details/121346483) 132 | 133 | 134 | > Github：[https://github.com/Whiffe/yolov5-deepsort](https://github.com/Whiffe/yolov5-deepsort) 135 | 136 | 遵循 GNU General Public License v3.0 协议，标明目标检测部分来源：https://github.com/ultralytics/yolov5/ 137 | 138 | 139 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/tracker.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | from __future__ import absolute_import 3 | import numpy as np 4 | from . import kalman_filter 5 | from . import linear_assignment 6 | from . import iou_matching 7 | from .track import Track 8 | 9 | 10 | class Tracker: 11 | 12 | def __init__(self, metric, max_iou_distance=0.7, max_age=70, n_init=3): 13 | self.metric = metric 14 | self.max_iou_distance = max_iou_distance 15 | self.max_age = max_age 16 | self.n_init = n_init 17 | 18 | self.kf = kalman_filter.KalmanFilter() 19 | self.tracks = [] 20 | self._next_id = 1 21 | 22 | def predict(self): 23 | """Propagate track state distributions one time step forward. 24 | 25 | This function should be called once every time step, before `update`. 26 | """ 27 | for track in self.tracks: 28 | track.predict(self.kf) 29 | 30 | def update(self, detections): 31 | """Perform measurement update and track management. 32 | 33 | Parameters 34 | ---------- 35 | detections : List[deep_sort.detection.Detection] 36 | A list of detections at the current time step. 37 | 38 | """ 39 | # Run matching cascade. 40 | matches, unmatched_tracks, unmatched_detections = \ 41 | self._match(detections) 42 | 43 | # Update track set. 44 | for track_idx, detection_idx in matches: 45 | self.tracks[track_idx].update( 46 | self.kf, detections[detection_idx]) 47 | for track_idx in unmatched_tracks: 48 | self.tracks[track_idx].mark_missed() 49 | for detection_idx in unmatched_detections: 50 | self._initiate_track(detections[detection_idx]) 51 | self.tracks = [t for t in self.tracks if not t.is_deleted()] 52 | 53 | # Update distance metric. 54 | active_targets = [t.track_id for t in self.tracks if t.is_confirmed()] 55 | features, targets = [], [] 56 | for track in self.tracks: 57 | if not track.is_confirmed(): 58 | continue 59 | features += track.features 60 | targets += [track.track_id for _ in track.features] 61 | track.features = [] 62 | self.metric.partial_fit( 63 | np.asarray(features), np.asarray(targets), active_targets) 64 | 65 | def _match(self, detections): 66 | 67 | def gated_metric(tracks, dets, track_indices, detection_indices): 68 | features = np.array([dets[i].feature for i in detection_indices]) 69 | targets = np.array([tracks[i].track_id for i in track_indices]) 70 | cost_matrix = self.metric.distance(features, targets) 71 | cost_matrix = linear_assignment.gate_cost_matrix( 72 | self.kf, cost_matrix, tracks, dets, track_indices, 73 | detection_indices) 74 | 75 | return cost_matrix 76 | 77 | # Split track set into confirmed and unconfirmed tracks. 78 | confirmed_tracks = [ 79 | i for i, t in enumerate(self.tracks) if t.is_confirmed()] 80 | unconfirmed_tracks = [ 81 | i for i, t in enumerate(self.tracks) if not t.is_confirmed()] 82 | 83 | # Associate confirmed tracks using appearance features. 84 | matches_a, unmatched_tracks_a, unmatched_detections = \ 85 | linear_assignment.matching_cascade( 86 | gated_metric, self.metric.matching_threshold, self.max_age, 87 | self.tracks, detections, confirmed_tracks) 88 | 89 | # Associate remaining tracks together with unconfirmed tracks using IOU. 90 | iou_track_candidates = unconfirmed_tracks + [ 91 | k for k in unmatched_tracks_a if 92 | self.tracks[k].time_since_update == 1] 93 | unmatched_tracks_a = [ 94 | k for k in unmatched_tracks_a if 95 | self.tracks[k].time_since_update != 1] 96 | matches_b, unmatched_tracks_b, unmatched_detections = \ 97 | linear_assignment.min_cost_matching( 98 | iou_matching.iou_cost, self.max_iou_distance, self.tracks, 99 | detections, iou_track_candidates, unmatched_detections) 100 | matches = matches_a + matches_b 101 | unmatched_tracks = list(set(unmatched_tracks_a + unmatched_tracks_b)) 102 | return matches, unmatched_tracks, unmatched_detections 103 | 104 | def _initiate_track(self, detection): 105 | mean, covariance = self.kf.initiate(detection.to_xyah()) 106 | self.tracks.append(Track( 107 | mean, detection.cls_, covariance, self._next_id, self.n_init, self.max_age, 108 | detection.feature)) 109 | self._next_id += 1 110 | -------------------------------------------------------------------------------- /deep_sort/utils/io.py: -------------------------------------------------------------------------------- 1 | import os 2 | from typing import Dict 3 | import numpy as np 4 | 5 | # from utils.log import get_logger 6 | 7 | 8 | def write_results(filename, results, data_type): 9 | if data_type == 'mot': 10 | save_format = '{frame},{id},{x1},{y1},{w},{h},-1,-1,-1,-1\n' 11 | elif data_type == 'kitti': 12 | save_format = '{frame} {id} pedestrian 0 0 -10 {x1} {y1} {x2} {y2} -10 -10 -10 -1000 -1000 -1000 -10\n' 13 | else: 14 | raise ValueError(data_type) 15 | 16 | with open(filename, 'w') as f: 17 | for frame_id, tlwhs, track_ids in results: 18 | if data_type == 'kitti': 19 | frame_id -= 1 20 | for tlwh, track_id in zip(tlwhs, track_ids): 21 | if track_id < 0: 22 | continue 23 | x1, y1, w, h = tlwh 24 | x2, y2 = x1 + w, y1 + h 25 | line = save_format.format(frame=frame_id, id=track_id, x1=x1, y1=y1, x2=x2, y2=y2, w=w, h=h) 26 | f.write(line) 27 | 28 | 29 | # def write_results(filename, results_dict: Dict, data_type: str): 30 | # if not filename: 31 | # return 32 | # path = os.path.dirname(filename) 33 | # if not os.path.exists(path): 34 | # os.makedirs(path) 35 | 36 | # if data_type in ('mot', 'mcmot', 'lab'): 37 | # save_format = '{frame},{id},{x1},{y1},{w},{h},1,-1,-1,-1\n' 38 | # elif data_type == 'kitti': 39 | # save_format = '{frame} {id} pedestrian -1 -1 -10 {x1} {y1} {x2} {y2} -1 -1 -1 -1000 -1000 -1000 -10 {score}\n' 40 | # else: 41 | # raise ValueError(data_type) 42 | 43 | # with open(filename, 'w') as f: 44 | # for frame_id, frame_data in results_dict.items(): 45 | # if data_type == 'kitti': 46 | # frame_id -= 1 47 | # for tlwh, track_id in frame_data: 48 | # if track_id < 0: 49 | # continue 50 | # x1, y1, w, h = tlwh 51 | # x2, y2 = x1 + w, y1 + h 52 | # line = save_format.format(frame=frame_id, id=track_id, x1=x1, y1=y1, x2=x2, y2=y2, w=w, h=h, score=1.0) 53 | # f.write(line) 54 | # logger.info('Save results to {}'.format(filename)) 55 | 56 | 57 | def read_results(filename, data_type: str, is_gt=False, is_ignore=False): 58 | if data_type in ('mot', 'lab'): 59 | read_fun = read_mot_results 60 | else: 61 | raise ValueError('Unknown data type: {}'.format(data_type)) 62 | 63 | return read_fun(filename, is_gt, is_ignore) 64 | 65 | 66 | """ 67 | labels={'ped', ... % 1 68 | 'person_on_vhcl', ... % 2 69 | 'car', ... % 3 70 | 'bicycle', ... % 4 71 | 'mbike', ... % 5 72 | 'non_mot_vhcl', ... % 6 73 | 'static_person', ... % 7 74 | 'distractor', ... % 8 75 | 'occluder', ... % 9 76 | 'occluder_on_grnd', ... %10 77 | 'occluder_full', ... % 11 78 | 'reflection', ... % 12 79 | 'crowd' ... % 13 80 | }; 81 | """ 82 | 83 | 84 | def read_mot_results(filename, is_gt, is_ignore): 85 | valid_labels = {1} 86 | ignore_labels = {2, 7, 8, 12} 87 | results_dict = dict() 88 | if os.path.isfile(filename): 89 | with open(filename, 'r') as f: 90 | for line in f.readlines(): 91 | linelist = line.split(',') 92 | if len(linelist) < 7: 93 | continue 94 | fid = int(linelist[0]) 95 | if fid < 1: 96 | continue 97 | results_dict.setdefault(fid, list()) 98 | 99 | if is_gt: 100 | if 'MOT16-' in filename or 'MOT17-' in filename: 101 | label = int(float(linelist[7])) 102 | mark = int(float(linelist[6])) 103 | if mark == 0 or label not in valid_labels: 104 | continue 105 | score = 1 106 | elif is_ignore: 107 | if 'MOT16-' in filename or 'MOT17-' in filename: 108 | label = int(float(linelist[7])) 109 | vis_ratio = float(linelist[8]) 110 | if label not in ignore_labels and vis_ratio >= 0: 111 | continue 112 | else: 113 | continue 114 | score = 1 115 | else: 116 | score = float(linelist[6]) 117 | 118 | tlwh = tuple(map(float, linelist[2:6])) 119 | target_id = int(linelist[1]) 120 | 121 | results_dict[fid].append((tlwh, target_id, score)) 122 | 123 | return results_dict 124 | 125 | 126 | def unzip_objs(objs): 127 | if len(objs) > 0: 128 | tlwhs, ids, scores = zip(*objs) 129 | else: 130 | tlwhs, ids, scores = [], [], [] 131 | tlwhs = np.asarray(tlwhs, dtype=float).reshape(-1, 4) 132 | 133 | return tlwhs, ids, scores -------------------------------------------------------------------------------- /utils/google_utils.py: -------------------------------------------------------------------------------- 1 | # Google utils: https://cloud.google.com/storage/docs/reference/libraries 2 | 3 | import os 4 | import platform 5 | import subprocess 6 | import time 7 | from pathlib import Path 8 | 9 | import requests 10 | import torch 11 | 12 | 13 | def gsutil_getsize(url=''): 14 | # gs://bucket/file size https://cloud.google.com/storage/docs/gsutil/commands/du 15 | s = subprocess.check_output(f'gsutil du {url}', shell=True).decode('utf-8') 16 | return eval(s.split(' ')[0]) if len(s) else 0 # bytes 17 | 18 | 19 | def attempt_download(file, repo='ultralytics/yolov5'): 20 | # Attempt file download if does not exist 21 | file = Path(str(file).strip().replace("'", '')) 22 | 23 | if not file.exists(): 24 | file.parent.mkdir(parents=True, exist_ok=True) # make parent dir (if required) 25 | try: 26 | response = requests.get(f'https://api.github.com/repos/{repo}/releases/latest').json() # github api 27 | assets = [x['name'] for x in response['assets']] # release assets, i.e. ['yolov5s.pt', 'yolov5m.pt', ...] 28 | tag = response['tag_name'] # i.e. 'v1.0' 29 | except: # fallback plan 30 | assets = ['yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt', 31 | 'yolov5s6.pt', 'yolov5m6.pt', 'yolov5l6.pt', 'yolov5x6.pt'] 32 | try: 33 | tag = subprocess.check_output('git tag', shell=True, stderr=subprocess.STDOUT).decode().split()[-1] 34 | except: 35 | tag = 'v5.0' # current release 36 | 37 | name = file.name 38 | if name in assets: 39 | msg = f'{file} missing, try downloading from https://github.com/{repo}/releases/' 40 | redundant = False # second download option 41 | try: # GitHub 42 | url = f'https://github.com/{repo}/releases/download/{tag}/{name}' 43 | print(f'Downloading {url} to {file}...') 44 | torch.hub.download_url_to_file(url, file) 45 | assert file.exists() and file.stat().st_size > 1E6 # check 46 | except Exception as e: # GCP 47 | print(f'Download error: {e}') 48 | assert redundant, 'No secondary mirror' 49 | url = f'https://storage.googleapis.com/{repo}/ckpt/{name}' 50 | print(f'Downloading {url} to {file}...') 51 | os.system(f"curl -L '{url}' -o '{file}' --retry 3 -C -") # curl download, retry and resume on fail 52 | finally: 53 | if not file.exists() or file.stat().st_size < 1E6: # check 54 | file.unlink(missing_ok=True) # remove partial downloads 55 | print(f'ERROR: Download failure: {msg}') 56 | print('') 57 | return 58 | 59 | 60 | def gdrive_download(id='16TiPfZj7htmTyhntwcZyEEAejOUxuT6m', file='tmp.zip'): 61 | # Downloads a file from Google Drive. from yolov5.utils.google_utils import *; gdrive_download() 62 | t = time.time() 63 | file = Path(file) 64 | cookie = Path('cookie') # gdrive cookie 65 | print(f'Downloading https://drive.google.com/uc?export=download&id={id} as {file}... ', end='') 66 | file.unlink(missing_ok=True) # remove existing file 67 | cookie.unlink(missing_ok=True) # remove existing cookie 68 | 69 | # Attempt file download 70 | out = "NUL" if platform.system() == "Windows" else "/dev/null" 71 | os.system(f'curl -c ./cookie -s -L "drive.google.com/uc?export=download&id={id}" > {out}') 72 | if os.path.exists('cookie'): # large file 73 | s = f'curl -Lb ./cookie "drive.google.com/uc?export=download&confirm={get_token()}&id={id}" -o {file}' 74 | else: # small file 75 | s = f'curl -s -L -o {file} "drive.google.com/uc?export=download&id={id}"' 76 | r = os.system(s) # execute, capture return 77 | cookie.unlink(missing_ok=True) # remove existing cookie 78 | 79 | # Error check 80 | if r != 0: 81 | file.unlink(missing_ok=True) # remove partial 82 | print('Download error ') # raise Exception('Download error') 83 | return r 84 | 85 | # Unzip if archive 86 | if file.suffix == '.zip': 87 | print('unzipping... ', end='') 88 | os.system(f'unzip -q {file}') # unzip 89 | file.unlink() # remove zip to free space 90 | 91 | print(f'Done ({time.time() - t:.1f}s)') 92 | return r 93 | 94 | 95 | def get_token(cookie="./cookie"): 96 | with open(cookie) as f: 97 | for line in f: 98 | if "download" in line: 99 | return line.split()[-1] 100 | return "" 101 | 102 | # def upload_blob(bucket_name, source_file_name, destination_blob_name): 103 | # # Uploads a file to a bucket 104 | # # https://cloud.google.com/storage/docs/uploading-objects#storage-upload-object-python 105 | # 106 | # storage_client = storage.Client() 107 | # bucket = storage_client.get_bucket(bucket_name) 108 | # blob = bucket.blob(destination_blob_name) 109 | # 110 | # blob.upload_from_filename(source_file_name) 111 | # 112 | # print('File {} uploaded to {}.'.format( 113 | # source_file_name, 114 | # destination_blob_name)) 115 | # 116 | # 117 | # def download_blob(bucket_name, source_blob_name, destination_file_name): 118 | # # Uploads a blob from a bucket 119 | # storage_client = storage.Client() 120 | # bucket = storage_client.get_bucket(bucket_name) 121 | # blob = bucket.blob(source_blob_name) 122 | # 123 | # blob.download_to_filename(destination_file_name) 124 | # 125 | # print('Blob {} downloaded to {}.'.format( 126 | # source_blob_name, 127 | # destination_file_name)) 128 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/track.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | 3 | 4 | class TrackState: 5 | """ 6 | Enumeration type for the single target track state. Newly created tracks are 7 | classified as `tentative` until enough evidence has been collected. Then, 8 | the track state is changed to `confirmed`. Tracks that are no longer alive 9 | are classified as `deleted` to mark them for removal from the set of active 10 | tracks. 11 | 12 | """ 13 | 14 | Tentative = 1 15 | Confirmed = 2 16 | Deleted = 3 17 | 18 | 19 | class Track: 20 | """ 21 | A single target track with state space `(x, y, a, h)` and associated 22 | velocities, where `(x, y)` is the center of the bounding box, `a` is the 23 | aspect ratio and `h` is the height. 24 | 25 | Parameters 26 | ---------- 27 | mean : ndarray 28 | Mean vector of the initial state distribution. 29 | covariance : ndarray 30 | Covariance matrix of the initial state distribution. 31 | track_id : int 32 | A unique track identifier. 33 | n_init : int 34 | Number of consecutive detections before the track is confirmed. The 35 | track state is set to `Deleted` if a miss occurs within the first 36 | `n_init` frames. 37 | max_age : int 38 | The maximum number of consecutive misses before the track state is 39 | set to `Deleted`. 40 | feature : Optional[ndarray] 41 | Feature vector of the detection this track originates from. If not None, 42 | this feature is added to the `features` cache. 43 | 44 | Attributes 45 | ---------- 46 | mean : ndarray 47 | Mean vector of the initial state distribution. 48 | covariance : ndarray 49 | Covariance matrix of the initial state distribution. 50 | track_id : int 51 | A unique track identifier. 52 | hits : int 53 | Total number of measurement updates. 54 | age : int 55 | Total number of frames since first occurance. 56 | time_since_update : int 57 | Total number of frames since last measurement update. 58 | state : TrackState 59 | The current track state. 60 | features : List[ndarray] 61 | A cache of features. On each measurement update, the associated feature 62 | vector is added to this list. 63 | 64 | """ 65 | 66 | def __init__(self, mean, cls_, covariance, track_id, n_init, max_age, 67 | feature=None): 68 | self.mean = mean 69 | self.cls_ = cls_ 70 | self.covariance = covariance 71 | self.track_id = track_id 72 | self.hits = 1 73 | self.age = 1 74 | self.time_since_update = 0 75 | 76 | self.state = TrackState.Tentative 77 | self.features = [] 78 | if feature is not None: 79 | self.features.append(feature) 80 | 81 | self._n_init = n_init 82 | self._max_age = max_age 83 | 84 | def to_tlwh(self): 85 | """Get current position in bounding box format `(top left x, top left y, 86 | width, height)`. 87 | 88 | Returns 89 | ------- 90 | ndarray 91 | The bounding box. 92 | 93 | """ 94 | ret = self.mean[:4].copy() 95 | ret[2] *= ret[3] 96 | ret[:2] -= ret[2:] / 2 97 | return ret 98 | 99 | def to_tlbr(self): 100 | """Get current position in bounding box format `(min x, miny, max x, 101 | max y)`. 102 | 103 | Returns 104 | ------- 105 | ndarray 106 | The bounding box. 107 | 108 | """ 109 | ret = self.to_tlwh() 110 | ret[2:] = ret[:2] + ret[2:] 111 | return ret 112 | 113 | def predict(self, kf): 114 | """Propagate the state distribution to the current time step using a 115 | Kalman filter prediction step. 116 | 117 | Parameters 118 | ---------- 119 | kf : kalman_filter.KalmanFilter 120 | The Kalman filter. 121 | 122 | """ 123 | self.mean, self.covariance = kf.predict(self.mean, self.covariance) 124 | self.age += 1 125 | self.time_since_update += 1 126 | 127 | def update(self, kf, detection): 128 | """Perform Kalman filter measurement update step and update the feature 129 | cache. 130 | 131 | Parameters 132 | ---------- 133 | kf : kalman_filter.KalmanFilter 134 | The Kalman filter. 135 | detection : Detection 136 | The associated detection. 137 | 138 | """ 139 | self.mean, self.covariance = kf.update( 140 | self.mean, self.covariance, detection.to_xyah()) 141 | self.features.append(detection.feature) 142 | self.cls_ = detection.cls_ 143 | 144 | self.hits += 1 145 | self.time_since_update = 0 146 | if self.state == TrackState.Tentative and self.hits >= self._n_init: 147 | self.state = TrackState.Confirmed 148 | 149 | def mark_missed(self): 150 | """Mark this track as missed (no association at the current time step). 151 | """ 152 | if self.state == TrackState.Tentative: 153 | self.state = TrackState.Deleted 154 | elif self.time_since_update > self._max_age: 155 | self.state = TrackState.Deleted 156 | 157 | def is_tentative(self): 158 | """Returns True if this track is tentative (unconfirmed). 159 | """ 160 | return self.state == TrackState.Tentative 161 | 162 | def is_confirmed(self): 163 | """Returns True if this track is confirmed.""" 164 | return self.state == TrackState.Confirmed 165 | 166 | def is_deleted(self): 167 | """Returns True if this track is dead and should be deleted.""" 168 | return self.state == TrackState.Deleted 169 | -------------------------------------------------------------------------------- /models/experimental.py: -------------------------------------------------------------------------------- 1 | # YOLOv5 experimental modules 2 | 3 | import numpy as np 4 | import torch 5 | import torch.nn as nn 6 | 7 | from models.common import Conv, DWConv 8 | from utils.google_utils import attempt_download 9 | 10 | 11 | class CrossConv(nn.Module): 12 | # Cross Convolution Downsample 13 | def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False): 14 | # ch_in, ch_out, kernel, stride, groups, expansion, shortcut 15 | super(CrossConv, self).__init__() 16 | c_ = int(c2 * e) # hidden channels 17 | self.cv1 = Conv(c1, c_, (1, k), (1, s)) 18 | self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g) 19 | self.add = shortcut and c1 == c2 20 | 21 | def forward(self, x): 22 | return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x)) 23 | 24 | 25 | class Sum(nn.Module): 26 | # Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070 27 | def __init__(self, n, weight=False): # n: number of inputs 28 | super(Sum, self).__init__() 29 | self.weight = weight # apply weights boolean 30 | self.iter = range(n - 1) # iter object 31 | if weight: 32 | self.w = nn.Parameter(-torch.arange(1., n) / 2, requires_grad=True) # layer weights 33 | 34 | def forward(self, x): 35 | y = x[0] # no weight 36 | if self.weight: 37 | w = torch.sigmoid(self.w) * 2 38 | for i in self.iter: 39 | y = y + x[i + 1] * w[i] 40 | else: 41 | for i in self.iter: 42 | y = y + x[i + 1] 43 | return y 44 | 45 | 46 | class GhostConv(nn.Module): 47 | # Ghost Convolution https://github.com/huawei-noah/ghostnet 48 | def __init__(self, c1, c2, k=1, s=1, g=1, act=True): # ch_in, ch_out, kernel, stride, groups 49 | super(GhostConv, self).__init__() 50 | c_ = c2 // 2 # hidden channels 51 | self.cv1 = Conv(c1, c_, k, s, None, g, act) 52 | self.cv2 = Conv(c_, c_, 5, 1, None, c_, act) 53 | 54 | def forward(self, x): 55 | y = self.cv1(x) 56 | return torch.cat([y, self.cv2(y)], 1) 57 | 58 | 59 | class GhostBottleneck(nn.Module): 60 | # Ghost Bottleneck https://github.com/huawei-noah/ghostnet 61 | def __init__(self, c1, c2, k=3, s=1): # ch_in, ch_out, kernel, stride 62 | super(GhostBottleneck, self).__init__() 63 | c_ = c2 // 2 64 | self.conv = nn.Sequential(GhostConv(c1, c_, 1, 1), # pw 65 | DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(), # dw 66 | GhostConv(c_, c2, 1, 1, act=False)) # pw-linear 67 | self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False), 68 | Conv(c1, c2, 1, 1, act=False)) if s == 2 else nn.Identity() 69 | 70 | def forward(self, x): 71 | return self.conv(x) + self.shortcut(x) 72 | 73 | 74 | class MixConv2d(nn.Module): 75 | # Mixed Depthwise Conv https://arxiv.org/abs/1907.09595 76 | def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True): 77 | super(MixConv2d, self).__init__() 78 | groups = len(k) 79 | if equal_ch: # equal c_ per group 80 | i = torch.linspace(0, groups - 1E-6, c2).floor() # c2 indices 81 | c_ = [(i == g).sum() for g in range(groups)] # intermediate channels 82 | else: # equal weight.numel() per group 83 | b = [c2] + [0] * groups 84 | a = np.eye(groups + 1, groups, k=-1) 85 | a -= np.roll(a, 1, axis=1) 86 | a *= np.array(k) ** 2 87 | a[0] = 1 88 | c_ = np.linalg.lstsq(a, b, rcond=None)[0].round() # solve for equal weight indices, ax = b 89 | 90 | self.m = nn.ModuleList([nn.Conv2d(c1, int(c_[g]), k[g], s, k[g] // 2, bias=False) for g in range(groups)]) 91 | self.bn = nn.BatchNorm2d(c2) 92 | self.act = nn.LeakyReLU(0.1, inplace=True) 93 | 94 | def forward(self, x): 95 | return x + self.act(self.bn(torch.cat([m(x) for m in self.m], 1))) 96 | 97 | 98 | class Ensemble(nn.ModuleList): 99 | # Ensemble of models 100 | def __init__(self): 101 | super(Ensemble, self).__init__() 102 | 103 | def forward(self, x, augment=False): 104 | y = [] 105 | for module in self: 106 | y.append(module(x, augment)[0]) 107 | # y = torch.stack(y).max(0)[0] # max ensemble 108 | # y = torch.stack(y).mean(0) # mean ensemble 109 | y = torch.cat(y, 1) # nms ensemble 110 | return y, None # inference, train output 111 | 112 | 113 | def attempt_load(weights, map_location=None, inplace=True): 114 | from models.yolo import Detect, Model 115 | 116 | # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a 117 | model = Ensemble() 118 | for w in weights if isinstance(weights, list) else [weights]: 119 | attempt_download(w) 120 | ckpt = torch.load(w, map_location=map_location) # load 121 | model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().fuse().eval()) # FP32 model 122 | 123 | # Compatibility updates 124 | for m in model.modules(): 125 | if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU, Detect, Model]: 126 | m.inplace = inplace # pytorch 1.7.0 compatibility 127 | elif type(m) is Conv: 128 | m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility 129 | 130 | if len(model) == 1: 131 | return model[-1] # return model 132 | else: 133 | print(f'Ensemble created with {weights}\n') 134 | for k in ['names']: 135 | setattr(model, k, getattr(model[-1], k)) 136 | model.stride = model[torch.argmax(torch.tensor([m.stride.max() for m in model])).int()].stride # max stride 137 | return model # return ensemble 138 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/nn_matching.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | import numpy as np 3 | 4 | 5 | def _pdist(a, b): 6 | """Compute pair-wise squared distance between points in `a` and `b`. 7 | 8 | Parameters 9 | ---------- 10 | a : array_like 11 | An NxM matrix of N samples of dimensionality M. 12 | b : array_like 13 | An LxM matrix of L samples of dimensionality M. 14 | 15 | Returns 16 | ------- 17 | ndarray 18 | Returns a matrix of size len(a), len(b) such that eleement (i, j) 19 | contains the squared distance between `a[i]` and `b[j]`. 20 | 21 | """ 22 | a, b = np.asarray(a), np.asarray(b) 23 | if len(a) == 0 or len(b) == 0: 24 | return np.zeros((len(a), len(b))) 25 | a2, b2 = np.square(a).sum(axis=1), np.square(b).sum(axis=1) 26 | r2 = -2. * np.dot(a, b.T) + a2[:, None] + b2[None, :] 27 | r2 = np.clip(r2, 0., float(np.inf)) 28 | return r2 29 | 30 | 31 | def _cosine_distance(a, b, data_is_normalized=False): 32 | """Compute pair-wise cosine distance between points in `a` and `b`. 33 | 34 | Parameters 35 | ---------- 36 | a : array_like 37 | An NxM matrix of N samples of dimensionality M. 38 | b : array_like 39 | An LxM matrix of L samples of dimensionality M. 40 | data_is_normalized : Optional[bool] 41 | If True, assumes rows in a and b are unit length vectors. 42 | Otherwise, a and b are explicitly normalized to lenght 1. 43 | 44 | Returns 45 | ------- 46 | ndarray 47 | Returns a matrix of size len(a), len(b) such that eleement (i, j) 48 | contains the squared distance between `a[i]` and `b[j]`. 49 | 50 | """ 51 | if not data_is_normalized: 52 | a = np.asarray(a) / np.linalg.norm(a, axis=1, keepdims=True) 53 | b = np.asarray(b) / np.linalg.norm(b, axis=1, keepdims=True) 54 | return 1. - np.dot(a, b.T) 55 | 56 | 57 | def _nn_euclidean_distance(x, y): 58 | """ Helper function for nearest neighbor distance metric (Euclidean). 59 | 60 | Parameters 61 | ---------- 62 | x : ndarray 63 | A matrix of N row-vectors (sample points). 64 | y : ndarray 65 | A matrix of M row-vectors (query points). 66 | 67 | Returns 68 | ------- 69 | ndarray 70 | A vector of length M that contains for each entry in `y` the 71 | smallest Euclidean distance to a sample in `x`. 72 | 73 | """ 74 | distances = _pdist(x, y) 75 | return np.maximum(0.0, distances.min(axis=0)) 76 | 77 | 78 | def _nn_cosine_distance(x, y): 79 | """ Helper function for nearest neighbor distance metric (cosine). 80 | 81 | Parameters 82 | ---------- 83 | x : ndarray 84 | A matrix of N row-vectors (sample points). 85 | y : ndarray 86 | A matrix of M row-vectors (query points). 87 | 88 | Returns 89 | ------- 90 | ndarray 91 | A vector of length M that contains for each entry in `y` the 92 | smallest cosine distance to a sample in `x`. 93 | 94 | """ 95 | distances = _cosine_distance(x, y) 96 | return distances.min(axis=0) 97 | 98 | 99 | class NearestNeighborDistanceMetric(object): 100 | """ 101 | A nearest neighbor distance metric that, for each target, returns 102 | the closest distance to any sample that has been observed so far. 103 | 104 | Parameters 105 | ---------- 106 | metric : str 107 | Either "euclidean" or "cosine". 108 | matching_threshold: float 109 | The matching threshold. Samples with larger distance are considered an 110 | invalid match. 111 | budget : Optional[int] 112 | If not None, fix samples per class to at most this number. Removes 113 | the oldest samples when the budget is reached. 114 | 115 | Attributes 116 | ---------- 117 | samples : Dict[int -> List[ndarray]] 118 | A dictionary that maps from target identities to the list of samples 119 | that have been observed so far. 120 | 121 | """ 122 | 123 | def __init__(self, metric, matching_threshold, budget=None): 124 | 125 | 126 | if metric == "euclidean": 127 | self._metric = _nn_euclidean_distance 128 | elif metric == "cosine": 129 | self._metric = _nn_cosine_distance 130 | else: 131 | raise ValueError( 132 | "Invalid metric; must be either 'euclidean' or 'cosine'") 133 | self.matching_threshold = matching_threshold 134 | self.budget = budget 135 | self.samples = {} 136 | 137 | def partial_fit(self, features, targets, active_targets): 138 | """Update the distance metric with new data. 139 | 140 | Parameters 141 | ---------- 142 | features : ndarray 143 | An NxM matrix of N features of dimensionality M. 144 | targets : ndarray 145 | An integer array of associated target identities. 146 | active_targets : List[int] 147 | A list of targets that are currently present in the scene. 148 | 149 | """ 150 | for feature, target in zip(features, targets): 151 | self.samples.setdefault(target, []).append(feature) 152 | if self.budget is not None: 153 | self.samples[target] = self.samples[target][-self.budget:] 154 | self.samples = {k: self.samples[k] for k in active_targets} 155 | 156 | def distance(self, features, targets): 157 | """Compute distance between features and targets. 158 | 159 | Parameters 160 | ---------- 161 | features : ndarray 162 | An NxM matrix of N features of dimensionality M. 163 | targets : List[int] 164 | A list of targets to match the given `features` against. 165 | 166 | Returns 167 | ------- 168 | ndarray 169 | Returns a cost matrix of shape len(targets), len(features), where 170 | element (i, j) contains the closest squared distance between 171 | `targets[i]` and `features[j]`. 172 | 173 | """ 174 | cost_matrix = np.zeros((len(targets), len(features))) 175 | for i, target in enumerate(targets): 176 | cost_matrix[i, :] = self._metric(self.samples[target], features) 177 | return cost_matrix 178 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/deep/train.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | import os 3 | import time 4 | 5 | import numpy as np 6 | import matplotlib.pyplot as plt 7 | import torch 8 | import torch.backends.cudnn as cudnn 9 | import torchvision 10 | 11 | from model import Net 12 | 13 | parser = argparse.ArgumentParser(description="Train on market1501") 14 | parser.add_argument("--data-dir",default='data',type=str) 15 | parser.add_argument("--no-cuda",action="store_true") 16 | parser.add_argument("--gpu-id",default=0,type=int) 17 | parser.add_argument("--lr",default=0.1, type=float) 18 | parser.add_argument("--interval",'-i',default=20,type=int) 19 | parser.add_argument('--resume', '-r',action='store_true') 20 | args = parser.parse_args() 21 | 22 | # device 23 | device = "cuda:{}".format(args.gpu_id) if torch.cuda.is_available() and not args.no_cuda else "cpu" 24 | if torch.cuda.is_available() and not args.no_cuda: 25 | cudnn.benchmark = True 26 | 27 | # data loading 28 | root = args.data_dir 29 | train_dir = os.path.join(root,"train") 30 | test_dir = os.path.join(root,"test") 31 | transform_train = torchvision.transforms.Compose([ 32 | torchvision.transforms.RandomCrop((128,64),padding=4), 33 | torchvision.transforms.RandomHorizontalFlip(), 34 | torchvision.transforms.ToTensor(), 35 | torchvision.transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) 36 | ]) 37 | transform_test = torchvision.transforms.Compose([ 38 | torchvision.transforms.Resize((128,64)), 39 | torchvision.transforms.ToTensor(), 40 | torchvision.transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]) 41 | ]) 42 | trainloader = torch.utils.data.DataLoader( 43 | torchvision.datasets.ImageFolder(train_dir, transform=transform_train), 44 | batch_size=64,shuffle=True 45 | ) 46 | testloader = torch.utils.data.DataLoader( 47 | torchvision.datasets.ImageFolder(test_dir, transform=transform_test), 48 | batch_size=64,shuffle=True 49 | ) 50 | num_classes = max(len(trainloader.dataset.classes), len(testloader.dataset.classes)) 51 | 52 | # net definition 53 | start_epoch = 0 54 | net = Net(num_classes=num_classes) 55 | if args.resume: 56 | assert os.path.isfile("./checkpoint/ckpt.t7"), "Error: no checkpoint file found!" 57 | print('Loading from checkpoint/ckpt.t7') 58 | checkpoint = torch.load("./checkpoint/ckpt.t7") 59 | # import ipdb; ipdb.set_trace() 60 | net_dict = checkpoint['net_dict'] 61 | net.load_state_dict(net_dict) 62 | best_acc = checkpoint['acc'] 63 | start_epoch = checkpoint['epoch'] 64 | net.to(device) 65 | 66 | # loss and optimizer 67 | criterion = torch.nn.CrossEntropyLoss() 68 | optimizer = torch.optim.SGD(net.parameters(), args.lr, momentum=0.9, weight_decay=5e-4) 69 | best_acc = 0. 70 | 71 | # train function for each epoch 72 | def train(epoch): 73 | print("\nEpoch : %d"%(epoch+1)) 74 | net.train() 75 | training_loss = 0. 76 | train_loss = 0. 77 | correct = 0 78 | total = 0 79 | interval = args.interval 80 | start = time.time() 81 | for idx, (inputs, labels) in enumerate(trainloader): 82 | # forward 83 | inputs,labels = inputs.to(device),labels.to(device) 84 | outputs = net(inputs) 85 | loss = criterion(outputs, labels) 86 | 87 | # backward 88 | optimizer.zero_grad() 89 | loss.backward() 90 | optimizer.step() 91 | 92 | # accumurating 93 | training_loss += loss.item() 94 | train_loss += loss.item() 95 | correct += outputs.max(dim=1)[1].eq(labels).sum().item() 96 | total += labels.size(0) 97 | 98 | # print 99 | if (idx+1)%interval == 0: 100 | end = time.time() 101 | print("[progress:{:.1f}%]time:{:.2f}s Loss:{:.5f} Correct:{}/{} Acc:{:.3f}%".format( 102 | 100.*(idx+1)/len(trainloader), end-start, training_loss/interval, correct, total, 100.*correct/total 103 | )) 104 | training_loss = 0. 105 | start = time.time() 106 | 107 | return train_loss/len(trainloader), 1.- correct/total 108 | 109 | def test(epoch): 110 | global best_acc 111 | net.eval() 112 | test_loss = 0. 113 | correct = 0 114 | total = 0 115 | start = time.time() 116 | with torch.no_grad(): 117 | for idx, (inputs, labels) in enumerate(testloader): 118 | inputs, labels = inputs.to(device), labels.to(device) 119 | outputs = net(inputs) 120 | loss = criterion(outputs, labels) 121 | 122 | test_loss += loss.item() 123 | correct += outputs.max(dim=1)[1].eq(labels).sum().item() 124 | total += labels.size(0) 125 | 126 | print("Testing ...") 127 | end = time.time() 128 | print("[progress:{:.1f}%]time:{:.2f}s Loss:{:.5f} Correct:{}/{} Acc:{:.3f}%".format( 129 | 100.*(idx+1)/len(testloader), end-start, test_loss/len(testloader), correct, total, 100.*correct/total 130 | )) 131 | 132 | # saving checkpoint 133 | acc = 100.*correct/total 134 | if acc > best_acc: 135 | best_acc = acc 136 | print("Saving parameters to checkpoint/ckpt.t7") 137 | checkpoint = { 138 | 'net_dict':net.state_dict(), 139 | 'acc':acc, 140 | 'epoch':epoch, 141 | } 142 | if not os.path.isdir('checkpoint'): 143 | os.mkdir('checkpoint') 144 | torch.save(checkpoint, './checkpoint/ckpt.t7') 145 | 146 | return test_loss/len(testloader), 1.- correct/total 147 | 148 | # plot figure 149 | x_epoch = [] 150 | record = {'train_loss':[], 'train_err':[], 'test_loss':[], 'test_err':[]} 151 | fig = plt.figure() 152 | ax0 = fig.add_subplot(121, title="loss") 153 | ax1 = fig.add_subplot(122, title="top1err") 154 | def draw_curve(epoch, train_loss, train_err, test_loss, test_err): 155 | global record 156 | record['train_loss'].append(train_loss) 157 | record['train_err'].append(train_err) 158 | record['test_loss'].append(test_loss) 159 | record['test_err'].append(test_err) 160 | 161 | x_epoch.append(epoch) 162 | ax0.plot(x_epoch, record['train_loss'], 'bo-', label='train') 163 | ax0.plot(x_epoch, record['test_loss'], 'ro-', label='val') 164 | ax1.plot(x_epoch, record['train_err'], 'bo-', label='train') 165 | ax1.plot(x_epoch, record['test_err'], 'ro-', label='val') 166 | if epoch == 0: 167 | ax0.legend() 168 | ax1.legend() 169 | fig.savefig("train.jpg") 170 | 171 | # lr decay 172 | def lr_decay(): 173 | global optimizer 174 | for params in optimizer.param_groups: 175 | params['lr'] *= 0.1 176 | lr = params['lr'] 177 | print("Learning rate adjusted to {}".format(lr)) 178 | 179 | def main(): 180 | for epoch in range(start_epoch, start_epoch+40): 181 | train_loss, train_err = train(epoch) 182 | test_loss, test_err = test(epoch) 183 | draw_curve(epoch, train_loss, train_err, test_loss, test_err) 184 | if (epoch+1)%20==0: 185 | lr_decay() 186 | 187 | 188 | if __name__ == '__main__': 189 | main() 190 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/linear_assignment.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | from __future__ import absolute_import 3 | import numpy as np 4 | # from sklearn.utils.linear_assignment_ import linear_assignment 5 | from scipy.optimize import linear_sum_assignment as linear_assignment 6 | from . import kalman_filter 7 | 8 | 9 | INFTY_COST = 1e+5 10 | 11 | 12 | def min_cost_matching( 13 | distance_metric, max_distance, tracks, detections, track_indices=None, 14 | detection_indices=None): 15 | if track_indices is None: 16 | track_indices = np.arange(len(tracks)) 17 | if detection_indices is None: 18 | detection_indices = np.arange(len(detections)) 19 | 20 | if len(detection_indices) == 0 or len(track_indices) == 0: 21 | return [], track_indices, detection_indices # Nothing to match. 22 | 23 | cost_matrix = distance_metric( 24 | tracks, detections, track_indices, detection_indices) 25 | cost_matrix[cost_matrix > max_distance] = max_distance + 1e-5 26 | 27 | row_indices, col_indices = linear_assignment(cost_matrix) 28 | 29 | matches, unmatched_tracks, unmatched_detections = [], [], [] 30 | for col, detection_idx in enumerate(detection_indices): 31 | if col not in col_indices: 32 | unmatched_detections.append(detection_idx) 33 | for row, track_idx in enumerate(track_indices): 34 | if row not in row_indices: 35 | unmatched_tracks.append(track_idx) 36 | for row, col in zip(row_indices, col_indices): 37 | track_idx = track_indices[row] 38 | detection_idx = detection_indices[col] 39 | if cost_matrix[row, col] > max_distance: 40 | unmatched_tracks.append(track_idx) 41 | unmatched_detections.append(detection_idx) 42 | else: 43 | matches.append((track_idx, detection_idx)) 44 | return matches, unmatched_tracks, unmatched_detections 45 | 46 | 47 | def matching_cascade( 48 | distance_metric, max_distance, cascade_depth, tracks, detections, 49 | track_indices=None, detection_indices=None): 50 | """Run matching cascade. 51 | 52 | Parameters 53 | ---------- 54 | distance_metric : Callable[List[Track], List[Detection], List[int], List[int]) -> ndarray 55 | The distance metric is given a list of tracks and detections as well as 56 | a list of N track indices and M detection indices. The metric should 57 | return the NxM dimensional cost matrix, where element (i, j) is the 58 | association cost between the i-th track in the given track indices and 59 | the j-th detection in the given detection indices. 60 | max_distance : float 61 | Gating threshold. Associations with cost larger than this value are 62 | disregarded. 63 | cascade_depth: int 64 | The cascade depth, should be se to the maximum track age. 65 | tracks : List[track.Track] 66 | A list of predicted tracks at the current time step. 67 | detections : List[detection.Detection] 68 | A list of detections at the current time step. 69 | track_indices : Optional[List[int]] 70 | List of track indices that maps rows in `cost_matrix` to tracks in 71 | `tracks` (see description above). Defaults to all tracks. 72 | detection_indices : Optional[List[int]] 73 | List of detection indices that maps columns in `cost_matrix` to 74 | detections in `detections` (see description above). Defaults to all 75 | detections. 76 | 77 | Returns 78 | ------- 79 | (List[(int, int)], List[int], List[int]) 80 | Returns a tuple with the following three entries: 81 | * A list of matched track and detection indices. 82 | * A list of unmatched track indices. 83 | * A list of unmatched detection indices. 84 | 85 | """ 86 | if track_indices is None: 87 | track_indices = list(range(len(tracks))) 88 | if detection_indices is None: 89 | detection_indices = list(range(len(detections))) 90 | 91 | unmatched_detections = detection_indices 92 | matches = [] 93 | for level in range(cascade_depth): 94 | if len(unmatched_detections) == 0: # No detections left 95 | break 96 | 97 | track_indices_l = [ 98 | k for k in track_indices 99 | if tracks[k].time_since_update == 1 + level 100 | ] 101 | if len(track_indices_l) == 0: # Nothing to match at this level 102 | continue 103 | 104 | matches_l, _, unmatched_detections = \ 105 | min_cost_matching( 106 | distance_metric, max_distance, tracks, detections, 107 | track_indices_l, unmatched_detections) 108 | matches += matches_l 109 | unmatched_tracks = list(set(track_indices) - set(k for k, _ in matches)) 110 | return matches, unmatched_tracks, unmatched_detections 111 | 112 | 113 | def gate_cost_matrix( 114 | kf, cost_matrix, tracks, detections, track_indices, detection_indices, 115 | gated_cost=INFTY_COST, only_position=False): 116 | """Invalidate infeasible entries in cost matrix based on the state 117 | distributions obtained by Kalman filtering. 118 | 119 | Parameters 120 | ---------- 121 | kf : The Kalman filter. 122 | cost_matrix : ndarray 123 | The NxM dimensional cost matrix, where N is the number of track indices 124 | and M is the number of detection indices, such that entry (i, j) is the 125 | association cost between `tracks[track_indices[i]]` and 126 | `detections[detection_indices[j]]`. 127 | tracks : List[track.Track] 128 | A list of predicted tracks at the current time step. 129 | detections : List[detection.Detection] 130 | A list of detections at the current time step. 131 | track_indices : List[int] 132 | List of track indices that maps rows in `cost_matrix` to tracks in 133 | `tracks` (see description above). 134 | detection_indices : List[int] 135 | List of detection indices that maps columns in `cost_matrix` to 136 | detections in `detections` (see description above). 137 | gated_cost : Optional[float] 138 | Entries in the cost matrix corresponding to infeasible associations are 139 | set this value. Defaults to a very large value. 140 | only_position : Optional[bool] 141 | If True, only the x, y position of the state distribution is considered 142 | during gating. Defaults to False. 143 | 144 | Returns 145 | ------- 146 | ndarray 147 | Returns the modified cost matrix. 148 | 149 | """ 150 | gating_dim = 2 if only_position else 4 151 | gating_threshold = kalman_filter.chi2inv95[gating_dim] 152 | measurements = np.asarray( 153 | [detections[i].to_xyah() for i in detection_indices]) 154 | for row, track_idx in enumerate(track_indices): 155 | track = tracks[track_idx] 156 | gating_distance = kf.gating_distance( 157 | track.mean, track.covariance, measurements, only_position) 158 | cost_matrix[row, gating_distance > gating_threshold] = gated_cost 159 | return cost_matrix 160 | -------------------------------------------------------------------------------- /models/export.py: -------------------------------------------------------------------------------- 1 | """Exports a YOLOv5 *.pt model to TorchScript, ONNX, CoreML formats 2 | 3 | Usage: 4 | $ python path/to/models/export.py --weights yolov5s.pt --img 640 --batch 1 5 | """ 6 | 7 | import argparse 8 | import sys 9 | import time 10 | from pathlib import Path 11 | 12 | sys.path.append(Path(__file__).parent.parent.absolute().__str__()) # to run '$ python *.py' files in subdirectories 13 | 14 | import torch 15 | import torch.nn as nn 16 | from torch.utils.mobile_optimizer import optimize_for_mobile 17 | 18 | import models 19 | from models.experimental import attempt_load 20 | from utils.activations import Hardswish, SiLU 21 | from utils.general import colorstr, check_img_size, check_requirements, file_size, set_logging 22 | from utils.torch_utils import select_device 23 | 24 | if __name__ == '__main__': 25 | parser = argparse.ArgumentParser() 26 | parser.add_argument('--weights', type=str, default='./yolov5s.pt', help='weights path') 27 | parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='image size') # height, width 28 | parser.add_argument('--batch-size', type=int, default=1, help='batch size') 29 | parser.add_argument('--device', default='cpu', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') 30 | parser.add_argument('--include', nargs='+', default=['torchscript', 'onnx', 'coreml'], help='include formats') 31 | parser.add_argument('--half', action='store_true', help='FP16 half-precision export') 32 | parser.add_argument('--inplace', action='store_true', help='set YOLOv5 Detect() inplace=True') 33 | parser.add_argument('--train', action='store_true', help='model.train() mode') 34 | parser.add_argument('--optimize', action='store_true', help='optimize TorchScript for mobile') # TorchScript-only 35 | parser.add_argument('--dynamic', action='store_true', help='dynamic ONNX axes') # ONNX-only 36 | parser.add_argument('--simplify', action='store_true', help='simplify ONNX model') # ONNX-only 37 | parser.add_argument('--opset-version', type=int, default=12, help='ONNX opset version') # ONNX-only 38 | opt = parser.parse_args() 39 | opt.img_size *= 2 if len(opt.img_size) == 1 else 1 # expand 40 | opt.include = [x.lower() for x in opt.include] 41 | print(opt) 42 | set_logging() 43 | t = time.time() 44 | 45 | # Load PyTorch model 46 | device = select_device(opt.device) 47 | model = attempt_load(opt.weights, map_location=device) # load FP32 model 48 | labels = model.names 49 | 50 | # Checks 51 | gs = int(max(model.stride)) # grid size (max stride) 52 | opt.img_size = [check_img_size(x, gs) for x in opt.img_size] # verify img_size are gs-multiples 53 | assert not (opt.device.lower() == 'cpu' and opt.half), '--half only compatible with GPU export, i.e. use --device 0' 54 | 55 | # Input 56 | img = torch.zeros(opt.batch_size, 3, *opt.img_size).to(device) # image size(1,3,320,192) iDetection 57 | 58 | # Update model 59 | if opt.half: 60 | img, model = img.half(), model.half() # to FP16 61 | if opt.train: 62 | model.train() # training mode (no grid construction in Detect layer) 63 | for k, m in model.named_modules(): 64 | m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility 65 | if isinstance(m, models.common.Conv): # assign export-friendly activations 66 | if isinstance(m.act, nn.Hardswish): 67 | m.act = Hardswish() 68 | elif isinstance(m.act, nn.SiLU): 69 | m.act = SiLU() 70 | elif isinstance(m, models.yolo.Detect): 71 | m.inplace = opt.inplace 72 | m.onnx_dynamic = opt.dynamic 73 | # m.forward = m.forward_export # assign forward (optional) 74 | 75 | for _ in range(2): 76 | y = model(img) # dry runs 77 | print(f"\n{colorstr('PyTorch:')} starting from {opt.weights} ({file_size(opt.weights):.1f} MB)") 78 | 79 | # TorchScript export ----------------------------------------------------------------------------------------------- 80 | if 'torchscript' in opt.include or 'coreml' in opt.include: 81 | prefix = colorstr('TorchScript:') 82 | try: 83 | print(f'\n{prefix} starting export with torch {torch.__version__}...') 84 | f = opt.weights.replace('.pt', '.torchscript.pt') # filename 85 | ts = torch.jit.trace(model, img, strict=False) 86 | (optimize_for_mobile(ts) if opt.optimize else ts).save(f) 87 | print(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') 88 | except Exception as e: 89 | print(f'{prefix} export failure: {e}') 90 | 91 | # ONNX export ------------------------------------------------------------------------------------------------------ 92 | if 'onnx' in opt.include: 93 | prefix = colorstr('ONNX:') 94 | try: 95 | import onnx 96 | 97 | print(f'{prefix} starting export with onnx {onnx.__version__}...') 98 | f = opt.weights.replace('.pt', '.onnx') # filename 99 | torch.onnx.export(model, img, f, verbose=False, opset_version=opt.opset_version, input_names=['images'], 100 | dynamic_axes={'images': {0: 'batch', 2: 'height', 3: 'width'}, # size(1,3,640,640) 101 | 'output': {0: 'batch', 2: 'y', 3: 'x'}} if opt.dynamic else None) 102 | 103 | # Checks 104 | model_onnx = onnx.load(f) # load onnx model 105 | onnx.checker.check_model(model_onnx) # check onnx model 106 | # print(onnx.helper.printable_graph(model_onnx.graph)) # print 107 | 108 | # Simplify 109 | if opt.simplify: 110 | try: 111 | check_requirements(['onnx-simplifier']) 112 | import onnxsim 113 | 114 | print(f'{prefix} simplifying with onnx-simplifier {onnxsim.__version__}...') 115 | model_onnx, check = onnxsim.simplify( 116 | model_onnx, 117 | dynamic_input_shape=opt.dynamic, 118 | input_shapes={'images': list(img.shape)} if opt.dynamic else None) 119 | assert check, 'assert check failed' 120 | onnx.save(model_onnx, f) 121 | except Exception as e: 122 | print(f'{prefix} simplifier failure: {e}') 123 | print(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') 124 | except Exception as e: 125 | print(f'{prefix} export failure: {e}') 126 | 127 | # CoreML export ---------------------------------------------------------------------------------------------------- 128 | if 'coreml' in opt.include: 129 | prefix = colorstr('CoreML:') 130 | try: 131 | import coremltools as ct 132 | 133 | print(f'{prefix} starting export with coremltools {ct.__version__}...') 134 | assert opt.train, 'CoreML exports should be placed in model.train() mode with `python export.py --train`' 135 | model = ct.convert(ts, inputs=[ct.ImageType('image', shape=img.shape, scale=1 / 255.0, bias=[0, 0, 0])]) 136 | f = opt.weights.replace('.pt', '.mlmodel') # filename 137 | model.save(f) 138 | print(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') 139 | except Exception as e: 140 | print(f'{prefix} export failure: {e}') 141 | 142 | # Finish 143 | print(f'\nExport complete ({time.time() - t:.2f}s). Visualize with https://github.com/lutzroeder/netron.') 144 | -------------------------------------------------------------------------------- /utils/autoanchor.py: -------------------------------------------------------------------------------- 1 | # Auto-anchor utils 2 | 3 | import numpy as np 4 | import torch 5 | import yaml 6 | from tqdm import tqdm 7 | 8 | from utils.general import colorstr 9 | 10 | 11 | def check_anchor_order(m): 12 | # Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary 13 | a = m.anchor_grid.prod(-1).view(-1) # anchor area 14 | da = a[-1] - a[0] # delta a 15 | ds = m.stride[-1] - m.stride[0] # delta s 16 | if da.sign() != ds.sign(): # same order 17 | print('Reversing anchor order') 18 | m.anchors[:] = m.anchors.flip(0) 19 | m.anchor_grid[:] = m.anchor_grid.flip(0) 20 | 21 | 22 | def check_anchors(dataset, model, thr=4.0, imgsz=640): 23 | # Check anchor fit to data, recompute if necessary 24 | prefix = colorstr('autoanchor: ') 25 | print(f'\n{prefix}Analyzing anchors... ', end='') 26 | m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1] # Detect() 27 | shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True) 28 | scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1)) # augment scale 29 | wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float() # wh 30 | 31 | def metric(k): # compute metric 32 | r = wh[:, None] / k[None] 33 | x = torch.min(r, 1. / r).min(2)[0] # ratio metric 34 | best = x.max(1)[0] # best_x 35 | aat = (x > 1. / thr).float().sum(1).mean() # anchors above threshold 36 | bpr = (best > 1. / thr).float().mean() # best possible recall 37 | return bpr, aat 38 | 39 | anchors = m.anchor_grid.clone().cpu().view(-1, 2) # current anchors 40 | bpr, aat = metric(anchors) 41 | print(f'anchors/target = {aat:.2f}, Best Possible Recall (BPR) = {bpr:.4f}', end='') 42 | if bpr < 0.98: # threshold to recompute 43 | print('. Attempting to improve anchors, please wait...') 44 | na = m.anchor_grid.numel() // 2 # number of anchors 45 | try: 46 | anchors = kmean_anchors(dataset, n=na, img_size=imgsz, thr=thr, gen=1000, verbose=False) 47 | except Exception as e: 48 | print(f'{prefix}ERROR: {e}') 49 | new_bpr = metric(anchors)[0] 50 | if new_bpr > bpr: # replace anchors 51 | anchors = torch.tensor(anchors, device=m.anchors.device).type_as(m.anchors) 52 | m.anchor_grid[:] = anchors.clone().view_as(m.anchor_grid) # for inference 53 | m.anchors[:] = anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1) # loss 54 | check_anchor_order(m) 55 | print(f'{prefix}New anchors saved to model. Update model *.yaml to use these anchors in the future.') 56 | else: 57 | print(f'{prefix}Original anchors better than new anchors. Proceeding with original anchors.') 58 | print('') # newline 59 | 60 | 61 | def kmean_anchors(path='./data/coco128.yaml', n=9, img_size=640, thr=4.0, gen=1000, verbose=True): 62 | """ Creates kmeans-evolved anchors from training dataset 63 | 64 | Arguments: 65 | path: path to dataset *.yaml, or a loaded dataset 66 | n: number of anchors 67 | img_size: image size used for training 68 | thr: anchor-label wh ratio threshold hyperparameter hyp['anchor_t'] used for training, default=4.0 69 | gen: generations to evolve anchors using genetic algorithm 70 | verbose: print all results 71 | 72 | Return: 73 | k: kmeans evolved anchors 74 | 75 | Usage: 76 | from utils.autoanchor import *; _ = kmean_anchors() 77 | """ 78 | from scipy.cluster.vq import kmeans 79 | 80 | thr = 1. / thr 81 | prefix = colorstr('autoanchor: ') 82 | 83 | def metric(k, wh): # compute metrics 84 | r = wh[:, None] / k[None] 85 | x = torch.min(r, 1. / r).min(2)[0] # ratio metric 86 | # x = wh_iou(wh, torch.tensor(k)) # iou metric 87 | return x, x.max(1)[0] # x, best_x 88 | 89 | def anchor_fitness(k): # mutation fitness 90 | _, best = metric(torch.tensor(k, dtype=torch.float32), wh) 91 | return (best * (best > thr).float()).mean() # fitness 92 | 93 | def print_results(k): 94 | k = k[np.argsort(k.prod(1))] # sort small to large 95 | x, best = metric(k, wh0) 96 | bpr, aat = (best > thr).float().mean(), (x > thr).float().mean() * n # best possible recall, anch > thr 97 | print(f'{prefix}thr={thr:.2f}: {bpr:.4f} best possible recall, {aat:.2f} anchors past thr') 98 | print(f'{prefix}n={n}, img_size={img_size}, metric_all={x.mean():.3f}/{best.mean():.3f}-mean/best, ' 99 | f'past_thr={x[x > thr].mean():.3f}-mean: ', end='') 100 | for i, x in enumerate(k): 101 | print('%i,%i' % (round(x[0]), round(x[1])), end=', ' if i < len(k) - 1 else '\n') # use in *.cfg 102 | return k 103 | 104 | if isinstance(path, str): # *.yaml file 105 | with open(path) as f: 106 | data_dict = yaml.safe_load(f) # model dict 107 | from utils.datasets import LoadImagesAndLabels 108 | dataset = LoadImagesAndLabels(data_dict['train'], augment=True, rect=True) 109 | else: 110 | dataset = path # dataset 111 | 112 | # Get label wh 113 | shapes = img_size * dataset.shapes / dataset.shapes.max(1, keepdims=True) 114 | wh0 = np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)]) # wh 115 | 116 | # Filter 117 | i = (wh0 < 3.0).any(1).sum() 118 | if i: 119 | print(f'{prefix}WARNING: Extremely small objects found. {i} of {len(wh0)} labels are < 3 pixels in size.') 120 | wh = wh0[(wh0 >= 2.0).any(1)] # filter > 2 pixels 121 | # wh = wh * (np.random.rand(wh.shape[0], 1) * 0.9 + 0.1) # multiply by random scale 0-1 122 | 123 | # Kmeans calculation 124 | print(f'{prefix}Running kmeans for {n} anchors on {len(wh)} points...') 125 | s = wh.std(0) # sigmas for whitening 126 | k, dist = kmeans(wh / s, n, iter=30) # points, mean distance 127 | assert len(k) == n, print(f'{prefix}ERROR: scipy.cluster.vq.kmeans requested {n} points but returned only {len(k)}') 128 | k *= s 129 | wh = torch.tensor(wh, dtype=torch.float32) # filtered 130 | wh0 = torch.tensor(wh0, dtype=torch.float32) # unfiltered 131 | k = print_results(k) 132 | 133 | # Plot 134 | # k, d = [None] * 20, [None] * 20 135 | # for i in tqdm(range(1, 21)): 136 | # k[i-1], d[i-1] = kmeans(wh / s, i) # points, mean distance 137 | # fig, ax = plt.subplots(1, 2, figsize=(14, 7), tight_layout=True) 138 | # ax = ax.ravel() 139 | # ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.') 140 | # fig, ax = plt.subplots(1, 2, figsize=(14, 7)) # plot wh 141 | # ax[0].hist(wh[wh[:, 0]<100, 0],400) 142 | # ax[1].hist(wh[wh[:, 1]<100, 1],400) 143 | # fig.savefig('wh.png', dpi=200) 144 | 145 | # Evolve 146 | npr = np.random 147 | f, sh, mp, s = anchor_fitness(k), k.shape, 0.9, 0.1 # fitness, generations, mutation prob, sigma 148 | pbar = tqdm(range(gen), desc=f'{prefix}Evolving anchors with Genetic Algorithm:') # progress bar 149 | for _ in pbar: 150 | v = np.ones(sh) 151 | while (v == 1).all(): # mutate until a change occurs (prevent duplicates) 152 | v = ((npr.random(sh) < mp) * npr.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0) 153 | kg = (k.copy() * v).clip(min=2.0) 154 | fg = anchor_fitness(kg) 155 | if fg > f: 156 | f, k = fg, kg.copy() 157 | pbar.desc = f'{prefix}Evolving anchors with Genetic Algorithm: fitness = {f:.4f}' 158 | if verbose: 159 | print_results(k) 160 | 161 | return print_results(k) 162 | -------------------------------------------------------------------------------- /deep_sort/deep_sort/sort/kalman_filter.py: -------------------------------------------------------------------------------- 1 | # vim: expandtab:ts=4:sw=4 2 | import numpy as np 3 | import scipy.linalg 4 | 5 | 6 | """ 7 | Table for the 0.95 quantile of the chi-square distribution with N degrees of 8 | freedom (contains values for N=1, ..., 9). Taken from MATLAB/Octave's chi2inv 9 | function and used as Mahalanobis gating threshold. 10 | """ 11 | chi2inv95 = { 12 | 1: 3.8415, 13 | 2: 5.9915, 14 | 3: 7.8147, 15 | 4: 9.4877, 16 | 5: 11.070, 17 | 6: 12.592, 18 | 7: 14.067, 19 | 8: 15.507, 20 | 9: 16.919} 21 | 22 | 23 | class KalmanFilter(object): 24 | """ 25 | A simple Kalman filter for tracking bounding boxes in image space. 26 | 27 | The 8-dimensional state space 28 | 29 | x, y, a, h, vx, vy, va, vh 30 | 31 | contains the bounding box center position (x, y), aspect ratio a, height h, 32 | and their respective velocities. 33 | 34 | Object motion follows a constant velocity model. The bounding box location 35 | (x, y, a, h) is taken as direct observation of the state space (linear 36 | observation model). 37 | 38 | """ 39 | 40 | def __init__(self): 41 | ndim, dt = 4, 1. 42 | 43 | # Create Kalman filter model matrices. 44 | self._motion_mat = np.eye(2 * ndim, 2 * ndim) 45 | for i in range(ndim): 46 | self._motion_mat[i, ndim + i] = dt 47 | self._update_mat = np.eye(ndim, 2 * ndim) 48 | 49 | # Motion and observation uncertainty are chosen relative to the current 50 | # state estimate. These weights control the amount of uncertainty in 51 | # the model. This is a bit hacky. 52 | self._std_weight_position = 1. / 20 53 | self._std_weight_velocity = 1. / 160 54 | 55 | def initiate(self, measurement): 56 | """Create track from unassociated measurement. 57 | 58 | Parameters 59 | ---------- 60 | measurement : ndarray 61 | Bounding box coordinates (x, y, a, h) with center position (x, y), 62 | aspect ratio a, and height h. 63 | 64 | Returns 65 | ------- 66 | (ndarray, ndarray) 67 | Returns the mean vector (8 dimensional) and covariance matrix (8x8 68 | dimensional) of the new track. Unobserved velocities are initialized 69 | to 0 mean. 70 | 71 | """ 72 | mean_pos = measurement 73 | mean_vel = np.zeros_like(mean_pos) 74 | mean = np.r_[mean_pos, mean_vel] 75 | 76 | std = [ 77 | 2 * self._std_weight_position * measurement[3], 78 | 2 * self._std_weight_position * measurement[3], 79 | 1e-2, 80 | 2 * self._std_weight_position * measurement[3], 81 | 10 * self._std_weight_velocity * measurement[3], 82 | 10 * self._std_weight_velocity * measurement[3], 83 | 1e-5, 84 | 10 * self._std_weight_velocity * measurement[3]] 85 | covariance = np.diag(np.square(std)) 86 | return mean, covariance 87 | 88 | def predict(self, mean, covariance): 89 | """Run Kalman filter prediction step. 90 | 91 | Parameters 92 | ---------- 93 | mean : ndarray 94 | The 8 dimensional mean vector of the object state at the previous 95 | time step. 96 | covariance : ndarray 97 | The 8x8 dimensional covariance matrix of the object state at the 98 | previous time step. 99 | 100 | Returns 101 | ------- 102 | (ndarray, ndarray) 103 | Returns the mean vector and covariance matrix of the predicted 104 | state. Unobserved velocities are initialized to 0 mean. 105 | 106 | """ 107 | std_pos = [ 108 | self._std_weight_position * mean[3], 109 | self._std_weight_position * mean[3], 110 | 1e-2, 111 | self._std_weight_position * mean[3]] 112 | std_vel = [ 113 | self._std_weight_velocity * mean[3], 114 | self._std_weight_velocity * mean[3], 115 | 1e-5, 116 | self._std_weight_velocity * mean[3]] 117 | motion_cov = np.diag(np.square(np.r_[std_pos, std_vel])) 118 | 119 | mean = np.dot(self._motion_mat, mean) 120 | covariance = np.linalg.multi_dot(( 121 | self._motion_mat, covariance, self._motion_mat.T)) + motion_cov 122 | 123 | return mean, covariance 124 | 125 | def project(self, mean, covariance): 126 | """Project state distribution to measurement space. 127 | 128 | Parameters 129 | ---------- 130 | mean : ndarray 131 | The state's mean vector (8 dimensional array). 132 | covariance : ndarray 133 | The state's covariance matrix (8x8 dimensional). 134 | 135 | Returns 136 | ------- 137 | (ndarray, ndarray) 138 | Returns the projected mean and covariance matrix of the given state 139 | estimate. 140 | 141 | """ 142 | std = [ 143 | self._std_weight_position * mean[3], 144 | self._std_weight_position * mean[3], 145 | 1e-1, 146 | self._std_weight_position * mean[3]] 147 | innovation_cov = np.diag(np.square(std)) 148 | 149 | mean = np.dot(self._update_mat, mean) 150 | covariance = np.linalg.multi_dot(( 151 | self._update_mat, covariance, self._update_mat.T)) 152 | return mean, covariance + innovation_cov 153 | 154 | def update(self, mean, covariance, measurement): 155 | """Run Kalman filter correction step. 156 | 157 | Parameters 158 | ---------- 159 | mean : ndarray 160 | The predicted state's mean vector (8 dimensional). 161 | covariance : ndarray 162 | The state's covariance matrix (8x8 dimensional). 163 | measurement : ndarray 164 | The 4 dimensional measurement vector (x, y, a, h), where (x, y) 165 | is the center position, a the aspect ratio, and h the height of the 166 | bounding box. 167 | 168 | Returns 169 | ------- 170 | (ndarray, ndarray) 171 | Returns the measurement-corrected state distribution. 172 | 173 | """ 174 | projected_mean, projected_cov = self.project(mean, covariance) 175 | 176 | chol_factor, lower = scipy.linalg.cho_factor( 177 | projected_cov, lower=True, check_finite=False) 178 | kalman_gain = scipy.linalg.cho_solve( 179 | (chol_factor, lower), np.dot(covariance, self._update_mat.T).T, 180 | check_finite=False).T 181 | innovation = measurement - projected_mean 182 | 183 | new_mean = mean + np.dot(innovation, kalman_gain.T) 184 | new_covariance = covariance - np.linalg.multi_dot(( 185 | kalman_gain, projected_cov, kalman_gain.T)) 186 | return new_mean, new_covariance 187 | 188 | def gating_distance(self, mean, covariance, measurements, 189 | only_position=False): 190 | """Compute gating distance between state distribution and measurements. 191 | 192 | A suitable distance threshold can be obtained from `chi2inv95`. If 193 | `only_position` is False, the chi-square distribution has 4 degrees of 194 | freedom, otherwise 2. 195 | 196 | Parameters 197 | ---------- 198 | mean : ndarray 199 | Mean vector over the state distribution (8 dimensional). 200 | covariance : ndarray 201 | Covariance of the state distribution (8x8 dimensional). 202 | measurements : ndarray 203 | An Nx4 dimensional matrix of N measurements, each in 204 | format (x, y, a, h) where (x, y) is the bounding box center 205 | position, a the aspect ratio, and h the height. 206 | only_position : Optional[bool] 207 | If True, distance computation is done with respect to the bounding 208 | box center position only. 209 | 210 | Returns 211 | ------- 212 | ndarray 213 | Returns an array of length N, where the i-th element contains the 214 | squared Mahalanobis distance between (mean, covariance) and 215 | `measurements[i]`. 216 | 217 | """ 218 | mean, covariance = self.project(mean, covariance) 219 | if only_position: 220 | mean, covariance = mean[:2], covariance[:2, :2] 221 | measurements = measurements[:, :2] 222 | 223 | cholesky_factor = np.linalg.cholesky(covariance) 224 | d = measurements - mean 225 | z = scipy.linalg.solve_triangular( 226 | cholesky_factor, d.T, lower=True, check_finite=False, 227 | overwrite_b=True) 228 | squared_maha = np.sum(z * z, axis=0) 229 | return squared_maha 230 | -------------------------------------------------------------------------------- /utils/metrics.py: -------------------------------------------------------------------------------- 1 | # Model validation metrics 2 | 3 | from pathlib import Path 4 | 5 | import matplotlib.pyplot as plt 6 | import numpy as np 7 | import torch 8 | 9 | from . import general 10 | 11 | 12 | def fitness(x): 13 | # Model fitness as a weighted combination of metrics 14 | w = [0.0, 0.0, 0.1, 0.9] # weights for [P, R, mAP@0.5, mAP@0.5:0.95] 15 | return (x[:, :4] * w).sum(1) 16 | 17 | 18 | def ap_per_class(tp, conf, pred_cls, target_cls, plot=False, save_dir='.', names=()): 19 | """ Compute the average precision, given the recall and precision curves. 20 | Source: https://github.com/rafaelpadilla/Object-Detection-Metrics. 21 | # Arguments 22 | tp: True positives (nparray, nx1 or nx10). 23 | conf: Objectness value from 0-1 (nparray). 24 | pred_cls: Predicted object classes (nparray). 25 | target_cls: True object classes (nparray). 26 | plot: Plot precision-recall curve at mAP@0.5 27 | save_dir: Plot save directory 28 | # Returns 29 | The average precision as computed in py-faster-rcnn. 30 | """ 31 | 32 | # Sort by objectness 33 | i = np.argsort(-conf) 34 | tp, conf, pred_cls = tp[i], conf[i], pred_cls[i] 35 | 36 | # Find unique classes 37 | unique_classes = np.unique(target_cls) 38 | nc = unique_classes.shape[0] # number of classes, number of detections 39 | 40 | # Create Precision-Recall curve and compute AP for each class 41 | px, py = np.linspace(0, 1, 1000), [] # for plotting 42 | ap, p, r = np.zeros((nc, tp.shape[1])), np.zeros((nc, 1000)), np.zeros((nc, 1000)) 43 | for ci, c in enumerate(unique_classes): 44 | i = pred_cls == c 45 | n_l = (target_cls == c).sum() # number of labels 46 | n_p = i.sum() # number of predictions 47 | 48 | if n_p == 0 or n_l == 0: 49 | continue 50 | else: 51 | # Accumulate FPs and TPs 52 | fpc = (1 - tp[i]).cumsum(0) 53 | tpc = tp[i].cumsum(0) 54 | 55 | # Recall 56 | recall = tpc / (n_l + 1e-16) # recall curve 57 | r[ci] = np.interp(-px, -conf[i], recall[:, 0], left=0) # negative x, xp because xp decreases 58 | 59 | # Precision 60 | precision = tpc / (tpc + fpc) # precision curve 61 | p[ci] = np.interp(-px, -conf[i], precision[:, 0], left=1) # p at pr_score 62 | 63 | # AP from recall-precision curve 64 | for j in range(tp.shape[1]): 65 | ap[ci, j], mpre, mrec = compute_ap(recall[:, j], precision[:, j]) 66 | if plot and j == 0: 67 | py.append(np.interp(px, mrec, mpre)) # precision at mAP@0.5 68 | 69 | # Compute F1 (harmonic mean of precision and recall) 70 | f1 = 2 * p * r / (p + r + 1e-16) 71 | if plot: 72 | plot_pr_curve(px, py, ap, Path(save_dir) / 'PR_curve.png', names) 73 | plot_mc_curve(px, f1, Path(save_dir) / 'F1_curve.png', names, ylabel='F1') 74 | plot_mc_curve(px, p, Path(save_dir) / 'P_curve.png', names, ylabel='Precision') 75 | plot_mc_curve(px, r, Path(save_dir) / 'R_curve.png', names, ylabel='Recall') 76 | 77 | i = f1.mean(0).argmax() # max F1 index 78 | return p[:, i], r[:, i], ap, f1[:, i], unique_classes.astype('int32') 79 | 80 | 81 | def compute_ap(recall, precision): 82 | """ Compute the average precision, given the recall and precision curves 83 | # Arguments 84 | recall: The recall curve (list) 85 | precision: The precision curve (list) 86 | # Returns 87 | Average precision, precision curve, recall curve 88 | """ 89 | 90 | # Append sentinel values to beginning and end 91 | mrec = np.concatenate(([0.], recall, [recall[-1] + 0.01])) 92 | mpre = np.concatenate(([1.], precision, [0.])) 93 | 94 | # Compute the precision envelope 95 | mpre = np.flip(np.maximum.accumulate(np.flip(mpre))) 96 | 97 | # Integrate area under curve 98 | method = 'interp' # methods: 'continuous', 'interp' 99 | if method == 'interp': 100 | x = np.linspace(0, 1, 101) # 101-point interp (COCO) 101 | ap = np.trapz(np.interp(x, mrec, mpre), x) # integrate 102 | else: # 'continuous' 103 | i = np.where(mrec[1:] != mrec[:-1])[0] # points where x axis (recall) changes 104 | ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1]) # area under curve 105 | 106 | return ap, mpre, mrec 107 | 108 | 109 | class ConfusionMatrix: 110 | # Updated version of https://github.com/kaanakan/object_detection_confusion_matrix 111 | def __init__(self, nc, conf=0.25, iou_thres=0.45): 112 | self.matrix = np.zeros((nc + 1, nc + 1)) 113 | self.nc = nc # number of classes 114 | self.conf = conf 115 | self.iou_thres = iou_thres 116 | 117 | def process_batch(self, detections, labels): 118 | """ 119 | Return intersection-over-union (Jaccard index) of boxes. 120 | Both sets of boxes are expected to be in (x1, y1, x2, y2) format. 121 | Arguments: 122 | detections (Array[N, 6]), x1, y1, x2, y2, conf, class 123 | labels (Array[M, 5]), class, x1, y1, x2, y2 124 | Returns: 125 | None, updates confusion matrix accordingly 126 | """ 127 | detections = detections[detections[:, 4] > self.conf] 128 | gt_classes = labels[:, 0].int() 129 | detection_classes = detections[:, 5].int() 130 | iou = general.box_iou(labels[:, 1:], detections[:, :4]) 131 | 132 | x = torch.where(iou > self.iou_thres) 133 | if x[0].shape[0]: 134 | matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy() 135 | if x[0].shape[0] > 1: 136 | matches = matches[matches[:, 2].argsort()[::-1]] 137 | matches = matches[np.unique(matches[:, 1], return_index=True)[1]] 138 | matches = matches[matches[:, 2].argsort()[::-1]] 139 | matches = matches[np.unique(matches[:, 0], return_index=True)[1]] 140 | else: 141 | matches = np.zeros((0, 3)) 142 | 143 | n = matches.shape[0] > 0 144 | m0, m1, _ = matches.transpose().astype(np.int16) 145 | for i, gc in enumerate(gt_classes): 146 | j = m0 == i 147 | if n and sum(j) == 1: 148 | self.matrix[detection_classes[m1[j]], gc] += 1 # correct 149 | else: 150 | self.matrix[self.nc, gc] += 1 # background FP 151 | 152 | if n: 153 | for i, dc in enumerate(detection_classes): 154 | if not any(m1 == i): 155 | self.matrix[dc, self.nc] += 1 # background FN 156 | 157 | def matrix(self): 158 | return self.matrix 159 | 160 | def plot(self, save_dir='', names=()): 161 | try: 162 | import seaborn as sn 163 | 164 | array = self.matrix / (self.matrix.sum(0).reshape(1, self.nc + 1) + 1E-6) # normalize 165 | array[array < 0.005] = np.nan # don't annotate (would appear as 0.00) 166 | 167 | fig = plt.figure(figsize=(12, 9), tight_layout=True) 168 | sn.set(font_scale=1.0 if self.nc < 50 else 0.8) # for label size 169 | labels = (0 < len(names) < 99) and len(names) == self.nc # apply names to ticklabels 170 | sn.heatmap(array, annot=self.nc < 30, annot_kws={"size": 8}, cmap='Blues', fmt='.2f', square=True, 171 | xticklabels=names + ['background FP'] if labels else "auto", 172 | yticklabels=names + ['background FN'] if labels else "auto").set_facecolor((1, 1, 1)) 173 | fig.axes[0].set_xlabel('True') 174 | fig.axes[0].set_ylabel('Predicted') 175 | fig.savefig(Path(save_dir) / 'confusion_matrix.png', dpi=250) 176 | except Exception as e: 177 | pass 178 | 179 | def print(self): 180 | for i in range(self.nc + 1): 181 | print(' '.join(map(str, self.matrix[i]))) 182 | 183 | 184 | # Plots ---------------------------------------------------------------------------------------------------------------- 185 | 186 | def plot_pr_curve(px, py, ap, save_dir='pr_curve.png', names=()): 187 | # Precision-recall curve 188 | fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True) 189 | py = np.stack(py, axis=1) 190 | 191 | if 0 < len(names) < 21: # display per-class legend if < 21 classes 192 | for i, y in enumerate(py.T): 193 | ax.plot(px, y, linewidth=1, label=f'{names[i]} {ap[i, 0]:.3f}') # plot(recall, precision) 194 | else: 195 | ax.plot(px, py, linewidth=1, color='grey') # plot(recall, precision) 196 | 197 | ax.plot(px, py.mean(1), linewidth=3, color='blue', label='all classes %.3f mAP@0.5' % ap[:, 0].mean()) 198 | ax.set_xlabel('Recall') 199 | ax.set_ylabel('Precision') 200 | ax.set_xlim(0, 1) 201 | ax.set_ylim(0, 1) 202 | plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left") 203 | fig.savefig(Path(save_dir), dpi=250) 204 | 205 | 206 | def plot_mc_curve(px, py, save_dir='mc_curve.png', names=(), xlabel='Confidence', ylabel='Metric'): 207 | # Metric-confidence curve 208 | fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True) 209 | 210 | if 0 < len(names) < 21: # display per-class legend if < 21 classes 211 | for i, y in enumerate(py): 212 | ax.plot(px, y, linewidth=1, label=f'{names[i]}') # plot(confidence, metric) 213 | else: 214 | ax.plot(px, py.T, linewidth=1, color='grey') # plot(confidence, metric) 215 | 216 | y = py.mean(0) 217 | ax.plot(px, y, linewidth=3, color='blue', label=f'all classes {y.max():.2f} at {px[y.argmax()]:.3f}') 218 | ax.set_xlabel(xlabel) 219 | ax.set_ylabel(ylabel) 220 | ax.set_xlim(0, 1) 221 | ax.set_ylim(0, 1) 222 | plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left") 223 | fig.savefig(Path(save_dir), dpi=250) 224 | -------------------------------------------------------------------------------- /utils/loss.py: -------------------------------------------------------------------------------- 1 | # Loss functions 2 | 3 | import torch 4 | import torch.nn as nn 5 | 6 | from utils.general import bbox_iou 7 | from utils.torch_utils import is_parallel 8 | 9 | 10 | def smooth_BCE(eps=0.1): # https://github.com/ultralytics/yolov3/issues/238#issuecomment-598028441 11 | # return positive, negative label smoothing BCE targets 12 | return 1.0 - 0.5 * eps, 0.5 * eps 13 | 14 | 15 | class BCEBlurWithLogitsLoss(nn.Module): 16 | # BCEwithLogitLoss() with reduced missing label effects. 17 | def __init__(self, alpha=0.05): 18 | super(BCEBlurWithLogitsLoss, self).__init__() 19 | self.loss_fcn = nn.BCEWithLogitsLoss(reduction='none') # must be nn.BCEWithLogitsLoss() 20 | self.alpha = alpha 21 | 22 | def forward(self, pred, true): 23 | loss = self.loss_fcn(pred, true) 24 | pred = torch.sigmoid(pred) # prob from logits 25 | dx = pred - true # reduce only missing label effects 26 | # dx = (pred - true).abs() # reduce missing label and false label effects 27 | alpha_factor = 1 - torch.exp((dx - 1) / (self.alpha + 1e-4)) 28 | loss *= alpha_factor 29 | return loss.mean() 30 | 31 | 32 | class FocalLoss(nn.Module): 33 | # Wraps focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5) 34 | def __init__(self, loss_fcn, gamma=1.5, alpha=0.25): 35 | super(FocalLoss, self).__init__() 36 | self.loss_fcn = loss_fcn # must be nn.BCEWithLogitsLoss() 37 | self.gamma = gamma 38 | self.alpha = alpha 39 | self.reduction = loss_fcn.reduction 40 | self.loss_fcn.reduction = 'none' # required to apply FL to each element 41 | 42 | def forward(self, pred, true): 43 | loss = self.loss_fcn(pred, true) 44 | # p_t = torch.exp(-loss) 45 | # loss *= self.alpha * (1.000001 - p_t) ** self.gamma # non-zero power for gradient stability 46 | 47 | # TF implementation https://github.com/tensorflow/addons/blob/v0.7.1/tensorflow_addons/losses/focal_loss.py 48 | pred_prob = torch.sigmoid(pred) # prob from logits 49 | p_t = true * pred_prob + (1 - true) * (1 - pred_prob) 50 | alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha) 51 | modulating_factor = (1.0 - p_t) ** self.gamma 52 | loss *= alpha_factor * modulating_factor 53 | 54 | if self.reduction == 'mean': 55 | return loss.mean() 56 | elif self.reduction == 'sum': 57 | return loss.sum() 58 | else: # 'none' 59 | return loss 60 | 61 | 62 | class QFocalLoss(nn.Module): 63 | # Wraps Quality focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5) 64 | def __init__(self, loss_fcn, gamma=1.5, alpha=0.25): 65 | super(QFocalLoss, self).__init__() 66 | self.loss_fcn = loss_fcn # must be nn.BCEWithLogitsLoss() 67 | self.gamma = gamma 68 | self.alpha = alpha 69 | self.reduction = loss_fcn.reduction 70 | self.loss_fcn.reduction = 'none' # required to apply FL to each element 71 | 72 | def forward(self, pred, true): 73 | loss = self.loss_fcn(pred, true) 74 | 75 | pred_prob = torch.sigmoid(pred) # prob from logits 76 | alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha) 77 | modulating_factor = torch.abs(true - pred_prob) ** self.gamma 78 | loss *= alpha_factor * modulating_factor 79 | 80 | if self.reduction == 'mean': 81 | return loss.mean() 82 | elif self.reduction == 'sum': 83 | return loss.sum() 84 | else: # 'none' 85 | return loss 86 | 87 | 88 | class ComputeLoss: 89 | # Compute losses 90 | def __init__(self, model, autobalance=False): 91 | super(ComputeLoss, self).__init__() 92 | device = next(model.parameters()).device # get model device 93 | h = model.hyp # hyperparameters 94 | 95 | # Define criteria 96 | BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device)) 97 | BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device)) 98 | 99 | # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3 100 | self.cp, self.cn = smooth_BCE(eps=h.get('label_smoothing', 0.0)) # positive, negative BCE targets 101 | 102 | # Focal loss 103 | g = h['fl_gamma'] # focal loss gamma 104 | if g > 0: 105 | BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g) 106 | 107 | det = model.module.model[-1] if is_parallel(model) else model.model[-1] # Detect() module 108 | self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.25, 0.06, .02]) # P3-P7 109 | self.ssi = list(det.stride).index(16) if autobalance else 0 # stride 16 index 110 | self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, model.gr, h, autobalance 111 | for k in 'na', 'nc', 'nl', 'anchors': 112 | setattr(self, k, getattr(det, k)) 113 | 114 | def __call__(self, p, targets): # predictions, targets, model 115 | device = targets.device 116 | lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device) 117 | tcls, tbox, indices, anchors = self.build_targets(p, targets) # targets 118 | 119 | # Losses 120 | for i, pi in enumerate(p): # layer index, layer predictions 121 | b, a, gj, gi = indices[i] # image, anchor, gridy, gridx 122 | tobj = torch.zeros_like(pi[..., 0], device=device) # target obj 123 | 124 | n = b.shape[0] # number of targets 125 | if n: 126 | ps = pi[b, a, gj, gi] # prediction subset corresponding to targets 127 | 128 | # Regression 129 | pxy = ps[:, :2].sigmoid() * 2. - 0.5 130 | pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i] 131 | pbox = torch.cat((pxy, pwh), 1) # predicted box 132 | iou = bbox_iou(pbox.T, tbox[i], x1y1x2y2=False, CIoU=True) # iou(prediction, target) 133 | lbox += (1.0 - iou).mean() # iou loss 134 | 135 | # Objectness 136 | tobj[b, a, gj, gi] = (1.0 - self.gr) + self.gr * iou.detach().clamp(0).type(tobj.dtype) # iou ratio 137 | 138 | # Classification 139 | if self.nc > 1: # cls loss (only if multiple classes) 140 | t = torch.full_like(ps[:, 5:], self.cn, device=device) # targets 141 | t[range(n), tcls[i]] = self.cp 142 | lcls += self.BCEcls(ps[:, 5:], t) # BCE 143 | 144 | # Append targets to text file 145 | # with open('targets.txt', 'a') as file: 146 | # [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)] 147 | 148 | obji = self.BCEobj(pi[..., 4], tobj) 149 | lobj += obji * self.balance[i] # obj loss 150 | if self.autobalance: 151 | self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item() 152 | 153 | if self.autobalance: 154 | self.balance = [x / self.balance[self.ssi] for x in self.balance] 155 | lbox *= self.hyp['box'] 156 | lobj *= self.hyp['obj'] 157 | lcls *= self.hyp['cls'] 158 | bs = tobj.shape[0] # batch size 159 | 160 | loss = lbox + lobj + lcls 161 | return loss * bs, torch.cat((lbox, lobj, lcls, loss)).detach() 162 | 163 | def build_targets(self, p, targets): 164 | # Build targets for compute_loss(), input targets(image,class,x,y,w,h) 165 | na, nt = self.na, targets.shape[0] # number of anchors, targets 166 | tcls, tbox, indices, anch = [], [], [], [] 167 | gain = torch.ones(7, device=targets.device) # normalized to gridspace gain 168 | ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt) # same as .repeat_interleave(nt) 169 | targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2) # append anchor indices 170 | 171 | g = 0.5 # bias 172 | off = torch.tensor([[0, 0], 173 | [1, 0], [0, 1], [-1, 0], [0, -1], # j,k,l,m 174 | # [1, 1], [1, -1], [-1, 1], [-1, -1], # jk,jm,lk,lm 175 | ], device=targets.device).float() * g # offsets 176 | 177 | for i in range(self.nl): 178 | anchors = self.anchors[i] 179 | gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]] # xyxy gain 180 | 181 | # Match targets to anchors 182 | t = targets * gain 183 | if nt: 184 | # Matches 185 | r = t[:, :, 4:6] / anchors[:, None] # wh ratio 186 | j = torch.max(r, 1. / r).max(2)[0] < self.hyp['anchor_t'] # compare 187 | # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t'] # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2)) 188 | t = t[j] # filter 189 | 190 | # Offsets 191 | gxy = t[:, 2:4] # grid xy 192 | gxi = gain[[2, 3]] - gxy # inverse 193 | j, k = ((gxy % 1. < g) & (gxy > 1.)).T 194 | l, m = ((gxi % 1. < g) & (gxi > 1.)).T 195 | j = torch.stack((torch.ones_like(j), j, k, l, m)) 196 | t = t.repeat((5, 1, 1))[j] 197 | offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j] 198 | else: 199 | t = targets[0] 200 | offsets = 0 201 | 202 | # Define 203 | b, c = t[:, :2].long().T # image, class 204 | gxy = t[:, 2:4] # grid xy 205 | gwh = t[:, 4:6] # grid wh 206 | gij = (gxy - offsets).long() 207 | gi, gj = gij.T # grid xy indices 208 | 209 | # Append 210 | a = t[:, 6].long() # anchor indices 211 | indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1))) # image, anchor, grid indices 212 | tbox.append(torch.cat((gxy - gij, gwh), 1)) # box 213 | anch.append(anchors[a]) # anchors 214 | tcls.append(c) # class 215 | 216 | return tcls, tbox, indices, anch 217 | -------------------------------------------------------------------------------- /deep_sort/utils/json_logger.py: -------------------------------------------------------------------------------- 1 | """ 2 | References: 3 | https://medium.com/analytics-vidhya/creating-a-custom-logging-mechanism-for-real-time-object-detection-using-tdd-4ca2cfcd0a2f 4 | """ 5 | import json 6 | from os import makedirs 7 | from os.path import exists, join 8 | from datetime import datetime 9 | 10 | 11 | class JsonMeta(object): 12 | HOURS = 3 13 | MINUTES = 59 14 | SECONDS = 59 15 | PATH_TO_SAVE = 'LOGS' 16 | DEFAULT_FILE_NAME = 'remaining' 17 | 18 | 19 | class BaseJsonLogger(object): 20 | """ 21 | This is the base class that returns __dict__ of its own 22 | it also returns the dicts of objects in the attributes that are list instances 23 | 24 | """ 25 | 26 | def dic(self): 27 | # returns dicts of objects 28 | out = {} 29 | for k, v in self.__dict__.items(): 30 | if hasattr(v, 'dic'): 31 | out[k] = v.dic() 32 | elif isinstance(v, list): 33 | out[k] = self.list(v) 34 | else: 35 | out[k] = v 36 | return out 37 | 38 | @staticmethod 39 | def list(values): 40 | # applies the dic method on items in the list 41 | return [v.dic() if hasattr(v, 'dic') else v for v in values] 42 | 43 | 44 | class Label(BaseJsonLogger): 45 | """ 46 | For each bounding box there are various categories with confidences. Label class keeps track of that information. 47 | """ 48 | 49 | def __init__(self, category: str, confidence: float): 50 | self.category = category 51 | self.confidence = confidence 52 | 53 | 54 | class Bbox(BaseJsonLogger): 55 | """ 56 | This module stores the information for each frame and use them in JsonParser 57 | Attributes: 58 | labels (list): List of label module. 59 | top (int): 60 | left (int): 61 | width (int): 62 | height (int): 63 | 64 | Args: 65 | bbox_id (float): 66 | top (int): 67 | left (int): 68 | width (int): 69 | height (int): 70 | 71 | References: 72 | Check Label module for better understanding. 73 | 74 | 75 | """ 76 | 77 | def __init__(self, bbox_id, top, left, width, height): 78 | self.labels = [] 79 | self.bbox_id = bbox_id 80 | self.top = top 81 | self.left = left 82 | self.width = width 83 | self.height = height 84 | 85 | def add_label(self, category, confidence): 86 | # adds category and confidence only if top_k is not exceeded. 87 | self.labels.append(Label(category, confidence)) 88 | 89 | def labels_full(self, value): 90 | return len(self.labels) == value 91 | 92 | 93 | class Frame(BaseJsonLogger): 94 | """ 95 | This module stores the information for each frame and use them in JsonParser 96 | Attributes: 97 | timestamp (float): The elapsed time of captured frame 98 | frame_id (int): The frame number of the captured video 99 | bboxes (list of Bbox objects): Stores the list of bbox objects. 100 | 101 | References: 102 | Check Bbox class for better information 103 | 104 | Args: 105 | timestamp (float): 106 | frame_id (int): 107 | 108 | """ 109 | 110 | def __init__(self, frame_id: int, timestamp: float = None): 111 | self.frame_id = frame_id 112 | self.timestamp = timestamp 113 | self.bboxes = [] 114 | 115 | def add_bbox(self, bbox_id: int, top: int, left: int, width: int, height: int): 116 | bboxes_ids = [bbox.bbox_id for bbox in self.bboxes] 117 | if bbox_id not in bboxes_ids: 118 | self.bboxes.append(Bbox(bbox_id, top, left, width, height)) 119 | else: 120 | raise ValueError("Frame with id: {} already has a Bbox with id: {}".format(self.frame_id, bbox_id)) 121 | 122 | def add_label_to_bbox(self, bbox_id: int, category: str, confidence: float): 123 | bboxes = {bbox.id: bbox for bbox in self.bboxes} 124 | if bbox_id in bboxes.keys(): 125 | res = bboxes.get(bbox_id) 126 | res.add_label(category, confidence) 127 | else: 128 | raise ValueError('the bbox with id: {} does not exists!'.format(bbox_id)) 129 | 130 | 131 | class BboxToJsonLogger(BaseJsonLogger): 132 | """ 133 | ُ This module is designed to automate the task of logging jsons. An example json is used 134 | to show the contents of json file shortly 135 | Example: 136 | { 137 | "video_details": { 138 | "frame_width": 1920, 139 | "frame_height": 1080, 140 | "frame_rate": 20, 141 | "video_name": "/home/gpu/codes/MSD/pedestrian_2/project/public/camera1.avi" 142 | }, 143 | "frames": [ 144 | { 145 | "frame_id": 329, 146 | "timestamp": 3365.1254 147 | "bboxes": [ 148 | { 149 | "labels": [ 150 | { 151 | "category": "pedestrian", 152 | "confidence": 0.9 153 | } 154 | ], 155 | "bbox_id": 0, 156 | "top": 1257, 157 | "left": 138, 158 | "width": 68, 159 | "height": 109 160 | } 161 | ] 162 | }], 163 | 164 | Attributes: 165 | frames (dict): It's a dictionary that maps each frame_id to json attributes. 166 | video_details (dict): information about video file. 167 | top_k_labels (int): shows the allowed number of labels 168 | start_time (datetime object): we use it to automate the json output by time. 169 | 170 | Args: 171 | top_k_labels (int): shows the allowed number of labels 172 | 173 | """ 174 | 175 | def __init__(self, top_k_labels: int = 1): 176 | self.frames = {} 177 | self.video_details = self.video_details = dict(frame_width=None, frame_height=None, frame_rate=None, 178 | video_name=None) 179 | self.top_k_labels = top_k_labels 180 | self.start_time = datetime.now() 181 | 182 | def set_top_k(self, value): 183 | self.top_k_labels = value 184 | 185 | def frame_exists(self, frame_id: int) -> bool: 186 | """ 187 | Args: 188 | frame_id (int): 189 | 190 | Returns: 191 | bool: true if frame_id is recognized 192 | """ 193 | return frame_id in self.frames.keys() 194 | 195 | def add_frame(self, frame_id: int, timestamp: float = None) -> None: 196 | """ 197 | Args: 198 | frame_id (int): 199 | timestamp (float): opencv captured frame time property 200 | 201 | Raises: 202 | ValueError: if frame_id would not exist in class frames attribute 203 | 204 | Returns: 205 | None 206 | 207 | """ 208 | if not self.frame_exists(frame_id): 209 | self.frames[frame_id] = Frame(frame_id, timestamp) 210 | else: 211 | raise ValueError("Frame id: {} already exists".format(frame_id)) 212 | 213 | def bbox_exists(self, frame_id: int, bbox_id: int) -> bool: 214 | """ 215 | Args: 216 | frame_id: 217 | bbox_id: 218 | 219 | Returns: 220 | bool: if bbox exists in frame bboxes list 221 | """ 222 | bboxes = [] 223 | if self.frame_exists(frame_id=frame_id): 224 | bboxes = [bbox.bbox_id for bbox in self.frames[frame_id].bboxes] 225 | return bbox_id in bboxes 226 | 227 | def find_bbox(self, frame_id: int, bbox_id: int): 228 | """ 229 | 230 | Args: 231 | frame_id: 232 | bbox_id: 233 | 234 | Returns: 235 | bbox_id (int): 236 | 237 | Raises: 238 | ValueError: if bbox_id does not exist in the bbox list of specific frame. 239 | """ 240 | if not self.bbox_exists(frame_id, bbox_id): 241 | raise ValueError("frame with id: {} does not contain bbox with id: {}".format(frame_id, bbox_id)) 242 | bboxes = {bbox.bbox_id: bbox for bbox in self.frames[frame_id].bboxes} 243 | return bboxes.get(bbox_id) 244 | 245 | def add_bbox_to_frame(self, frame_id: int, bbox_id: int, top: int, left: int, width: int, height: int) -> None: 246 | """ 247 | 248 | Args: 249 | frame_id (int): 250 | bbox_id (int): 251 | top (int): 252 | left (int): 253 | width (int): 254 | height (int): 255 | 256 | Returns: 257 | None 258 | 259 | Raises: 260 | ValueError: if bbox_id already exist in frame information with frame_id 261 | ValueError: if frame_id does not exist in frames attribute 262 | """ 263 | if self.frame_exists(frame_id): 264 | frame = self.frames[frame_id] 265 | if not self.bbox_exists(frame_id, bbox_id): 266 | frame.add_bbox(bbox_id, top, left, width, height) 267 | else: 268 | raise ValueError( 269 | "frame with frame_id: {} already contains the bbox with id: {} ".format(frame_id, bbox_id)) 270 | else: 271 | raise ValueError("frame with frame_id: {} does not exist".format(frame_id)) 272 | 273 | def add_label_to_bbox(self, frame_id: int, bbox_id: int, category: str, confidence: float): 274 | """ 275 | Args: 276 | frame_id: 277 | bbox_id: 278 | category: 279 | confidence: the confidence value returned from yolo detection 280 | 281 | Returns: 282 | None 283 | 284 | Raises: 285 | ValueError: if labels quota (top_k_labels) exceeds. 286 | """ 287 | bbox = self.find_bbox(frame_id, bbox_id) 288 | if not bbox.labels_full(self.top_k_labels): 289 | bbox.add_label(category, confidence) 290 | else: 291 | raise ValueError("labels in frame_id: {}, bbox_id: {} is fulled".format(frame_id, bbox_id)) 292 | 293 | def add_video_details(self, frame_width: int = None, frame_height: int = None, frame_rate: int = None, 294 | video_name: str = None): 295 | self.video_details['frame_width'] = frame_width 296 | self.video_details['frame_height'] = frame_height 297 | self.video_details['frame_rate'] = frame_rate 298 | self.video_details['video_name'] = video_name 299 | 300 | def output(self): 301 | output = {'video_details': self.video_details} 302 | result = list(self.frames.values()) 303 | output['frames'] = [item.dic() for item in result] 304 | return output 305 | 306 | def json_output(self, output_name): 307 | """ 308 | Args: 309 | output_name: 310 | 311 | Returns: 312 | None 313 | 314 | Notes: 315 | It creates the json output with `output_name` name. 316 | """ 317 | if not output_name.endswith('.json'): 318 | output_name += '.json' 319 | with open(output_name, 'w') as file: 320 | json.dump(self.output(), file) 321 | file.close() 322 | 323 | def set_start(self): 324 | self.start_time = datetime.now() 325 | 326 | def schedule_output_by_time(self, output_dir=JsonMeta.PATH_TO_SAVE, hours: int = 0, minutes: int = 0, 327 | seconds: int = 60) -> None: 328 | """ 329 | Notes: 330 | Creates folder and then periodically stores the jsons on that address. 331 | 332 | Args: 333 | output_dir (str): the directory where output files will be stored 334 | hours (int): 335 | minutes (int): 336 | seconds (int): 337 | 338 | Returns: 339 | None 340 | 341 | """ 342 | end = datetime.now() 343 | interval = 0 344 | interval += abs(min([hours, JsonMeta.HOURS]) * 3600) 345 | interval += abs(min([minutes, JsonMeta.MINUTES]) * 60) 346 | interval += abs(min([seconds, JsonMeta.SECONDS])) 347 | diff = (end - self.start_time).seconds 348 | 349 | if diff > interval: 350 | output_name = self.start_time.strftime('%Y-%m-%d %H-%M-%S') + '.json' 351 | if not exists(output_dir): 352 | makedirs(output_dir) 353 | output = join(output_dir, output_name) 354 | self.json_output(output_name=output) 355 | self.frames = {} 356 | self.start_time = datetime.now() 357 | 358 | def schedule_output_by_frames(self, frames_quota, frame_counter, output_dir=JsonMeta.PATH_TO_SAVE): 359 | """ 360 | saves as the number of frames quota increases higher. 361 | :param frames_quota: 362 | :param frame_counter: 363 | :param output_dir: 364 | :return: 365 | """ 366 | pass 367 | 368 | def flush(self, output_dir): 369 | """ 370 | Notes: 371 | We use this function to output jsons whenever possible. 372 | like the time that we exit the while loop of opencv. 373 | 374 | Args: 375 | output_dir: 376 | 377 | Returns: 378 | None 379 | 380 | """ 381 | filename = self.start_time.strftime('%Y-%m-%d %H-%M-%S') + '-remaining.json' 382 | output = join(output_dir, filename) 383 | self.json_output(output_name=output) 384 | --------------------------------------------------------------------------------