├── .idea ├── .gitignore ├── fast_yolov7_pytorch.iml ├── inspectionProfiles │ ├── Project_Default.xml │ └── profiles_settings.xml ├── modules.xml └── vcs.xml ├── README.md ├── detect.py ├── models ├── __init__.py ├── common.py ├── experimental.py ├── quant_yolo.py └── yolo.py ├── papers └── README.md ├── prune.py ├── requirements.txt ├── test.py ├── train.py └── utils ├── __init__.py ├── activations.py ├── add_nms.py ├── autoanchor.py ├── aws ├── __init__.py ├── mime.sh ├── resume.py └── userdata.sh ├── datasets.py ├── general.py ├── google_app_engine ├── Dockerfile ├── additional_requirements.txt └── app.yaml ├── google_utils.py ├── loss.py ├── metrics.py ├── plots.py ├── torch_utils.py └── wandb_logging ├── __init__.py ├── log_dataset.py └── wandb_utils.py /.idea/.gitignore: -------------------------------------------------------------------------------- 1 | # Default ignored files 2 | /shelf/ 3 | /workspace.xml 4 | -------------------------------------------------------------------------------- /.idea/fast_yolov7_pytorch.iml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 12 | -------------------------------------------------------------------------------- /.idea/inspectionProfiles/Project_Default.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 34 | -------------------------------------------------------------------------------- /.idea/inspectionProfiles/profiles_settings.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 6 | -------------------------------------------------------------------------------- /.idea/modules.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | -------------------------------------------------------------------------------- /.idea/vcs.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Fast_Yolov7_Pytorch🎉️🎉️🎉️ 2 | 3 | ***Use SOTA Pruning and Quant Algorithm to Build Your Faster Yolov7🚀️*** 4 | 5 | ## Guide 6 | 7 | * [Installation](#index1) 8 | * [Quick Start](#index2) 9 | * [References](#index3) 10 | * [Contact Me](#index4) 11 | 12 | ## Installation： 13 | 14 | ```commandline 15 | git clone https://github.com/ChristianYang37/fast_yolov7_pytorch.git 16 | ``` 17 | 18 | ```commandline 19 | pip install -r requirements.txt 20 | ``` 21 | 22 | If you can't install torch_pruning, please do as follow 23 | 24 | ```commandline 25 | git clone https://github.com/VainF/Torch-Pruning.git 26 | cd Torch-Pruning 27 | python setup.py install 28 | ``` 29 | 30 | For pytorch yolov7 state dicts, [click here](https://github.com/WongKinYiu/yolov7#transfer-learning) to download. 31 | 32 | ## Quick Start： 33 | 34 | ### Pruning & Training 35 | 36 | ```commandline 37 | python train.py --workers 8 --device 0 --batch-size 32 --data data/custom.yaml --img 640 640 --cfg cfg/training/yolov7-custom.yaml --weights 'yolov7_training.pt' --name yolov7-custom --hyp data/hyp.scratch.custom.yaml --sparsity 0.3 --num_epoch_to_prune 4 --prune_nore L2 38 | ``` 39 | 40 | if you want prune model without training, you can just set `epochs` = 0 41 | 42 | #### Opts 43 | 44 | `sparsity`: the sparsity of pruning 45 | 46 | `num_epoch_to_prune`: prune model after `num_epoch_to_prune` times finetune 47 | 48 | `prune_norm`: L1 or L2 49 | 50 | the code actually do prune as follows 51 | 52 | ````python 53 | for idx, epoch in enumerate(range(start_epoch, epochs)): 54 | if (idx + 1) % opt.num_epochs_to_prune: 55 | yolo_pruner.step(model, device) 56 | 57 | ```` 58 | 59 | So, for more efficient pruning, we suggest you set `num_batch_to_prune` big enough to make sure the model has fitted the data before you prune it, and also set `epochs` optimally. 60 | 61 | #### Some Results 62 | 63 | Prune `yolov7_training.pt` On COCO128.yaml (without finetune) 64 | 65 | 66 | | Sparsity | Macs | num_params | mAP@.5 | mAP@.0:.95 | 67 | | -------- | ---------- | ---------- | ------- | ---------- | 68 | | 0 | 6501867771 | 37622682 | 0.817 | 0.615 | 69 | | 0.005 | 6379356844 | 37115689 | 0.791 | 0.541 | 70 | | 0.007 | 6373571463 | 37033908 | 0.783 | 0.515 | 71 | | 0.01 | 6324846255 | 36735256 | 0.758 | 0.508 | 72 | | 0.02 | 6187011754 | 35974768 | 0.615 | 0.38 | 73 | | 0.05 | 5820065160 | 33891742 | 0.25 | 0.123 | 74 | | 0.1 | 5237469860 | 30417686 | 0.00056 | 0.000102 | 75 | 76 | Speed test on `GPU=A5000, batch_size=32` 77 | 78 | 79 | | Sparsity | batch 32 average time / s | 80 | | -------- | ------------------------- | 81 | | 0 | 0.055983 | 82 | | 0.005 | 0.044586 | 83 | | 0.01 | 0.044711 | 84 | | 0.05 | 0.043469 | 85 | | 0.1 | 0.041813 | 86 | | 0.2 | 0.037244 | 87 | | 0.5 | 0.023613 | 88 | | 0.7 | 0.024631 | 89 | 90 | ### Quantization 91 | 92 | ```commandline 93 | python train.py --workers 8 --device 0 --batch-size 32 --data data/custom.yaml --img 640 640 --cfg cfg/training/yolov7-custom.yaml --weights 'yolov7_training.pt' --name yolov7-custom --hyp data/hyp.scratch.custom.yaml --method static 94 | ``` 95 | 96 | #### Opts 97 | 98 | `method`: algorithm to quantify model, static or dynamic 99 | 100 | `deploy_device`: pytorch now support x86 and arm, is enabled for `method` == static only 101 | 102 | When you set `method` = dynamic, it require train data to make quantified model fit the distribution. 103 | 104 | ## References: 105 | 106 | 1. [WongKinYiu/yolov7: Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors (github.com)](https://github.com/WongKinYiu/yolov7) 107 | 2. [VainF/Torch-Pruning: [CVPR-2023] Towards Any Structural Pruning; LLaMA / CNNs / Transformers (github.com)](https://github.com/VainF/Torch-Pruning) 108 | 3. [PyTorch](https://pytorch.org/) 109 | 4. [ultralytics/yolov5: YOLOv5 🚀 in PyTorch > ONNX > CoreML > TFLite (github.com)](https://github.com/ultralytics/yolov5) 110 | 111 | ## Contact： 112 | 113 | This repository is for [AIRS](https://airs.cuhk.edu.cn/)'s project, the author is an undergraduate student at Sun Yat sen University. 114 | 115 | Email: christiannyang37@gmail.com 116 | -------------------------------------------------------------------------------- /detect.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | import time 3 | from pathlib import Path 4 | 5 | import cv2 6 | import torch 7 | import torch.backends.cudnn as cudnn 8 | from numpy import random 9 | 10 | from models.experimental import attempt_load 11 | from utils.datasets import LoadStreams, LoadImages 12 | from utils.general import check_img_size, check_requirements, check_imshow, non_max_suppression, apply_classifier, \ 13 | scale_coords, xyxy2xywh, strip_optimizer, set_logging, increment_path 14 | from utils.plots import plot_one_box 15 | from utils.torch_utils import select_device, load_classifier, time_synchronized, TracedModel 16 | 17 | 18 | def detect(save_img=False): 19 | source, weights, view_img, save_txt, imgsz, trace = opt.source, opt.weights, opt.view_img, opt.save_txt, opt.img_size, not opt.no_trace 20 | save_img = not opt.nosave and not source.endswith('.txt') # save inference images 21 | webcam = source.isnumeric() or source.endswith('.txt') or source.lower().startswith( 22 | ('rtsp://', 'rtmp://', 'http://', 'https://')) 23 | 24 | # Directories 25 | save_dir = Path(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok)) # increment run 26 | (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True) # make dir 27 | 28 | # Initialize 29 | set_logging() 30 | device = select_device(opt.device) 31 | half = device.type != 'cpu' # half precision only supported on CUDA 32 | 33 | # Load model 34 | model = attempt_load(weights, map_location=device) # load FP32 model 35 | stride = int(model.stride.max()) # model stride 36 | imgsz = check_img_size(imgsz, s=stride) # check img_size 37 | 38 | if trace: 39 | model = TracedModel(model, device, opt.img_size) 40 | 41 | if half: 42 | model.half() # to FP16 43 | 44 | # Second-stage classifier 45 | classify = False 46 | if classify: 47 | modelc = load_classifier(name='resnet101', n=2) # initialize 48 | modelc.load_state_dict(torch.load('weights/resnet101.pt', map_location=device)['model']).to(device).eval() 49 | 50 | # Set Dataloader 51 | vid_path, vid_writer = None, None 52 | if webcam: 53 | view_img = check_imshow() 54 | cudnn.benchmark = True # set True to speed up constant image size inference 55 | dataset = LoadStreams(source, img_size=imgsz, stride=stride) 56 | else: 57 | dataset = LoadImages(source, img_size=imgsz, stride=stride) 58 | 59 | # Get names and colors 60 | names = model.module.names if hasattr(model, 'module') else model.names 61 | colors = [[random.randint(0, 255) for _ in range(3)] for _ in names] 62 | 63 | # Run inference 64 | if device.type != 'cpu': 65 | model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters()))) # run once 66 | old_img_w = old_img_h = imgsz 67 | old_img_b = 1 68 | 69 | t0 = time.time() 70 | for path, img, im0s, vid_cap in dataset: 71 | img = torch.from_numpy(img).to(device) 72 | img = img.half() if half else img.float() # uint8 to fp16/32 73 | img /= 255.0 # 0 - 255 to 0.0 - 1.0 74 | if img.ndimension() == 3: 75 | img = img.unsqueeze(0) 76 | 77 | # Warmup 78 | if device.type != 'cpu' and (old_img_b != img.shape[0] or old_img_h != img.shape[2] or old_img_w != img.shape[3]): 79 | old_img_b = img.shape[0] 80 | old_img_h = img.shape[2] 81 | old_img_w = img.shape[3] 82 | for i in range(3): 83 | model(img, augment=opt.augment)[0] 84 | 85 | # Inference 86 | t1 = time_synchronized() 87 | with torch.no_grad(): # Calculating gradients would cause a GPU memory leak 88 | pred = model(img, augment=opt.augment)[0] 89 | t2 = time_synchronized() 90 | 91 | # Apply NMS 92 | pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres, classes=opt.classes, agnostic=opt.agnostic_nms) 93 | t3 = time_synchronized() 94 | 95 | # Apply Classifier 96 | if classify: 97 | pred = apply_classifier(pred, modelc, img, im0s) 98 | 99 | # Process detections 100 | for i, det in enumerate(pred): # detections per image 101 | if webcam: # batch_size >= 1 102 | p, s, im0, frame = path[i], '%g: ' % i, im0s[i].copy(), dataset.count 103 | else: 104 | p, s, im0, frame = path, '', im0s, getattr(dataset, 'frame', 0) 105 | 106 | p = Path(p) # to Path 107 | save_path = str(save_dir / p.name) # img.jpg 108 | txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}') # img.txt 109 | gn = torch.tensor(im0.shape)[[1, 0, 1, 0]] # normalization gain whwh 110 | if len(det): 111 | # Rescale boxes from img_size to im0 size 112 | det[:, :4] = scale_coords(img.shape[2:], det[:, :4], im0.shape).round() 113 | 114 | # Print results 115 | for c in det[:, -1].unique(): 116 | n = (det[:, -1] == c).sum() # detections per class 117 | s += f"{n} {names[int(c)]}{'s' * (n > 1)}, " # add to string 118 | 119 | # Write results 120 | for *xyxy, conf, cls in reversed(det): 121 | if save_txt: # Write to file 122 | xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh 123 | line = (cls, *xywh, conf) if opt.save_conf else (cls, *xywh) # label format 124 | with open(txt_path + '.txt', 'a') as f: 125 | f.write(('%g ' * len(line)).rstrip() % line + '\n') 126 | 127 | if save_img or view_img: # Add bbox to image 128 | label = f'{names[int(cls)]} {conf:.2f}' 129 | plot_one_box(xyxy, im0, label=label, color=colors[int(cls)], line_thickness=1) 130 | 131 | # Print time (inference + NMS) 132 | print(f'{s}Done. ({(1E3 * (t2 - t1)):.1f}ms) Inference, ({(1E3 * (t3 - t2)):.1f}ms) NMS') 133 | 134 | # Stream results 135 | if view_img: 136 | cv2.imshow(str(p), im0) 137 | cv2.waitKey(1) # 1 millisecond 138 | 139 | # Save results (image with detections) 140 | if save_img: 141 | if dataset.mode == 'image': 142 | cv2.imwrite(save_path, im0) 143 | print(f" The image with the result is saved in: {save_path}") 144 | else: # 'video' or 'stream' 145 | if vid_path != save_path: # new video 146 | vid_path = save_path 147 | if isinstance(vid_writer, cv2.VideoWriter): 148 | vid_writer.release() # release previous video writer 149 | if vid_cap: # video 150 | fps = vid_cap.get(cv2.CAP_PROP_FPS) 151 | w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH)) 152 | h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) 153 | else: # stream 154 | fps, w, h = 30, im0.shape[1], im0.shape[0] 155 | save_path += '.mp4' 156 | vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h)) 157 | vid_writer.write(im0) 158 | 159 | if save_txt or save_img: 160 | s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else '' 161 | #print(f"Results saved to {save_dir}{s}") 162 | 163 | print(f'Done. ({time.time() - t0:.3f}s)') 164 | 165 | 166 | if __name__ == '__main__': 167 | parser = argparse.ArgumentParser() 168 | parser.add_argument('--weights', nargs='+', type=str, default='yolov7.pt', help='model.pt path(s)') 169 | parser.add_argument('--source', type=str, default='inference/images', help='source') # file/folder, 0 for webcam 170 | parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)') 171 | parser.add_argument('--conf-thres', type=float, default=0.25, help='object confidence threshold') 172 | parser.add_argument('--iou-thres', type=float, default=0.45, help='IOU threshold for NMS') 173 | parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') 174 | parser.add_argument('--view-img', action='store_true', help='display results') 175 | parser.add_argument('--save-txt', action='store_true', help='save results to *.txt') 176 | parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels') 177 | parser.add_argument('--nosave', action='store_true', help='do not save images/videos') 178 | parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3') 179 | parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS') 180 | parser.add_argument('--augment', action='store_true', help='augmented inference') 181 | parser.add_argument('--update', action='store_true', help='update all models') 182 | parser.add_argument('--project', default='runs/detect', help='save results to project/name') 183 | parser.add_argument('--name', default='exp', help='save results to project/name') 184 | parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment') 185 | parser.add_argument('--no-trace', action='store_true', help='don`t trace model') 186 | opt = parser.parse_args() 187 | print(opt) 188 | #check_requirements(exclude=('pycocotools', 'thop')) 189 | 190 | with torch.no_grad(): 191 | if opt.update: # update all models (to fix SourceChangeWarning) 192 | for opt.weights in ['yolov7.pt']: 193 | detect() 194 | strip_optimizer(opt.weights) 195 | else: 196 | detect() 197 | -------------------------------------------------------------------------------- /models/__init__.py: -------------------------------------------------------------------------------- 1 | # init -------------------------------------------------------------------------------- /models/experimental.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import random 3 | import torch 4 | import torch.nn as nn 5 | 6 | from models.common import Conv, DWConv 7 | from utils.google_utils import attempt_download 8 | 9 | 10 | class CrossConv(nn.Module): 11 | # Cross Convolution Downsample 12 | def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False): 13 | # ch_in, ch_out, kernel, stride, groups, expansion, shortcut 14 | super(CrossConv, self).__init__() 15 | c_ = int(c2 * e) # hidden channels 16 | self.cv1 = Conv(c1, c_, (1, k), (1, s)) 17 | self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g) 18 | self.add = shortcut and c1 == c2 19 | 20 | def forward(self, x): 21 | return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x)) 22 | 23 | 24 | class Sum(nn.Module): 25 | # Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070 26 | def __init__(self, n, weight=False): # n: number of inputs 27 | super(Sum, self).__init__() 28 | self.weight = weight # apply weights boolean 29 | self.iter = range(n - 1) # iter object 30 | if weight: 31 | self.w = nn.Parameter(-torch.arange(1., n) / 2, requires_grad=True) # layer weights 32 | 33 | def forward(self, x): 34 | y = x[0] # no weight 35 | if self.weight: 36 | w = torch.sigmoid(self.w) * 2 37 | for i in self.iter: 38 | y = y + x[i + 1] * w[i] 39 | else: 40 | for i in self.iter: 41 | y = y + x[i + 1] 42 | return y 43 | 44 | 45 | class MixConv2d(nn.Module): 46 | # Mixed Depthwise Conv https://arxiv.org/abs/1907.09595 47 | def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True): 48 | super(MixConv2d, self).__init__() 49 | groups = len(k) 50 | if equal_ch: # equal c_ per group 51 | i = torch.linspace(0, groups - 1E-6, c2).floor() # c2 indices 52 | c_ = [(i == g).sum() for g in range(groups)] # intermediate channels 53 | else: # equal weight.numel() per group 54 | b = [c2] + [0] * groups 55 | a = np.eye(groups + 1, groups, k=-1) 56 | a -= np.roll(a, 1, axis=1) 57 | a *= np.array(k) ** 2 58 | a[0] = 1 59 | c_ = np.linalg.lstsq(a, b, rcond=None)[0].round() # solve for equal weight indices, ax = b 60 | 61 | self.m = nn.ModuleList([nn.Conv2d(c1, int(c_[g]), k[g], s, k[g] // 2, bias=False) for g in range(groups)]) 62 | self.bn = nn.BatchNorm2d(c2) 63 | self.act = nn.LeakyReLU(0.1, inplace=True) 64 | 65 | def forward(self, x): 66 | return x + self.act(self.bn(torch.cat([m(x) for m in self.m], 1))) 67 | 68 | 69 | class Ensemble(nn.ModuleList): 70 | # Ensemble of models 71 | def __init__(self): 72 | super(Ensemble, self).__init__() 73 | 74 | def forward(self, x, augment=False): 75 | y = [] 76 | for module in self: 77 | y.append(module(x, augment)[0]) 78 | # y = torch.stack(y).max(0)[0] # max ensemble 79 | # y = torch.stack(y).mean(0) # mean ensemble 80 | y = torch.cat(y, 1) # nms ensemble 81 | return y, None # inference, train output 82 | 83 | 84 | 85 | 86 | 87 | class ORT_NMS(torch.autograd.Function): 88 | '''ONNX-Runtime NMS operation''' 89 | @staticmethod 90 | def forward(ctx, 91 | boxes, 92 | scores, 93 | max_output_boxes_per_class=torch.tensor([100]), 94 | iou_threshold=torch.tensor([0.45]), 95 | score_threshold=torch.tensor([0.25])): 96 | device = boxes.device 97 | batch = scores.shape[0] 98 | num_det = random.randint(0, 100) 99 | batches = torch.randint(0, batch, (num_det,)).sort()[0].to(device) 100 | idxs = torch.arange(100, 100 + num_det).to(device) 101 | zeros = torch.zeros((num_det,), dtype=torch.int64).to(device) 102 | selected_indices = torch.cat([batches[None], zeros[None], idxs[None]], 0).T.contiguous() 103 | selected_indices = selected_indices.to(torch.int64) 104 | return selected_indices 105 | 106 | @staticmethod 107 | def symbolic(g, boxes, scores, max_output_boxes_per_class, iou_threshold, score_threshold): 108 | return g.op("NonMaxSuppression", boxes, scores, max_output_boxes_per_class, iou_threshold, score_threshold) 109 | 110 | 111 | class TRT_NMS(torch.autograd.Function): 112 | '''TensorRT NMS operation''' 113 | @staticmethod 114 | def forward( 115 | ctx, 116 | boxes, 117 | scores, 118 | background_class=-1, 119 | box_coding=1, 120 | iou_threshold=0.45, 121 | max_output_boxes=100, 122 | plugin_version="1", 123 | score_activation=0, 124 | score_threshold=0.25, 125 | ): 126 | batch_size, num_boxes, num_classes = scores.shape 127 | num_det = torch.randint(0, max_output_boxes, (batch_size, 1), dtype=torch.int32) 128 | det_boxes = torch.randn(batch_size, max_output_boxes, 4) 129 | det_scores = torch.randn(batch_size, max_output_boxes) 130 | det_classes = torch.randint(0, num_classes, (batch_size, max_output_boxes), dtype=torch.int32) 131 | return num_det, det_boxes, det_scores, det_classes 132 | 133 | @staticmethod 134 | def symbolic(g, 135 | boxes, 136 | scores, 137 | background_class=-1, 138 | box_coding=1, 139 | iou_threshold=0.45, 140 | max_output_boxes=100, 141 | plugin_version="1", 142 | score_activation=0, 143 | score_threshold=0.25): 144 | out = g.op("TRT::EfficientNMS_TRT", 145 | boxes, 146 | scores, 147 | background_class_i=background_class, 148 | box_coding_i=box_coding, 149 | iou_threshold_f=iou_threshold, 150 | max_output_boxes_i=max_output_boxes, 151 | plugin_version_s=plugin_version, 152 | score_activation_i=score_activation, 153 | score_threshold_f=score_threshold, 154 | outputs=4) 155 | nums, boxes, scores, classes = out 156 | return nums, boxes, scores, classes 157 | 158 | 159 | class ONNX_ORT(nn.Module): 160 | '''onnx module with ONNX-Runtime NMS operation.''' 161 | def __init__(self, max_obj=100, iou_thres=0.45, score_thres=0.25, max_wh=640, device=None, n_classes=80): 162 | super().__init__() 163 | self.device = device if device else torch.device("cpu") 164 | self.max_obj = torch.tensor([max_obj]).to(device) 165 | self.iou_threshold = torch.tensor([iou_thres]).to(device) 166 | self.score_threshold = torch.tensor([score_thres]).to(device) 167 | self.max_wh = max_wh # if max_wh != 0 : non-agnostic else : agnostic 168 | self.convert_matrix = torch.tensor([[1, 0, 1, 0], [0, 1, 0, 1], [-0.5, 0, 0.5, 0], [0, -0.5, 0, 0.5]], 169 | dtype=torch.float32, 170 | device=self.device) 171 | self.n_classes=n_classes 172 | 173 | def forward(self, x): 174 | boxes = x[:, :, :4] 175 | conf = x[:, :, 4:5] 176 | scores = x[:, :, 5:] 177 | if self.n_classes == 1: 178 | scores = conf # for models with one class, cls_loss is 0 and cls_conf is always 0.5, 179 | # so there is no need to multiplicate. 180 | else: 181 | scores *= conf # conf = obj_conf * cls_conf 182 | boxes @= self.convert_matrix 183 | max_score, category_id = scores.max(2, keepdim=True) 184 | dis = category_id.float() * self.max_wh 185 | nmsbox = boxes + dis 186 | max_score_tp = max_score.transpose(1, 2).contiguous() 187 | selected_indices = ORT_NMS.apply(nmsbox, max_score_tp, self.max_obj, self.iou_threshold, self.score_threshold) 188 | X, Y = selected_indices[:, 0], selected_indices[:, 2] 189 | selected_boxes = boxes[X, Y, :] 190 | selected_categories = category_id[X, Y, :].float() 191 | selected_scores = max_score[X, Y, :] 192 | X = X.unsqueeze(1).float() 193 | return torch.cat([X, selected_boxes, selected_categories, selected_scores], 1) 194 | 195 | class ONNX_TRT(nn.Module): 196 | '''onnx module with TensorRT NMS operation.''' 197 | def __init__(self, max_obj=100, iou_thres=0.45, score_thres=0.25, max_wh=None ,device=None, n_classes=80): 198 | super().__init__() 199 | assert max_wh is None 200 | self.device = device if device else torch.device('cpu') 201 | self.background_class = -1, 202 | self.box_coding = 1, 203 | self.iou_threshold = iou_thres 204 | self.max_obj = max_obj 205 | self.plugin_version = '1' 206 | self.score_activation = 0 207 | self.score_threshold = score_thres 208 | self.n_classes=n_classes 209 | 210 | def forward(self, x): 211 | boxes = x[:, :, :4] 212 | conf = x[:, :, 4:5] 213 | scores = x[:, :, 5:] 214 | if self.n_classes == 1: 215 | scores = conf # for models with one class, cls_loss is 0 and cls_conf is always 0.5, 216 | # so there is no need to multiplicate. 217 | else: 218 | scores *= conf # conf = obj_conf * cls_conf 219 | num_det, det_boxes, det_scores, det_classes = TRT_NMS.apply(boxes, scores, self.background_class, self.box_coding, 220 | self.iou_threshold, self.max_obj, 221 | self.plugin_version, self.score_activation, 222 | self.score_threshold) 223 | return num_det, det_boxes, det_scores, det_classes 224 | 225 | 226 | class End2End(nn.Module): 227 | '''export onnx or tensorrt model with NMS operation.''' 228 | def __init__(self, model, max_obj=100, iou_thres=0.45, score_thres=0.25, max_wh=None, device=None, n_classes=80): 229 | super().__init__() 230 | device = device if device else torch.device('cpu') 231 | assert isinstance(max_wh,(int)) or max_wh is None 232 | self.model = model.to(device) 233 | self.model.model[-1].end2end = True 234 | self.patch_model = ONNX_TRT if max_wh is None else ONNX_ORT 235 | self.end2end = self.patch_model(max_obj, iou_thres, score_thres, max_wh, device, n_classes) 236 | self.end2end.eval() 237 | 238 | def forward(self, x): 239 | x = self.model(x) 240 | x = self.end2end(x) 241 | return x 242 | 243 | 244 | 245 | 246 | 247 | def attempt_load(weights, map_location=None): 248 | # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a 249 | model = Ensemble() 250 | for w in weights if isinstance(weights, list) else [weights]: 251 | attempt_download(w) 252 | ckpt = torch.load(w, map_location=map_location) # load 253 | model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().fuse().eval()) # FP32 model 254 | 255 | # Compatibility updates 256 | for m in model.modules(): 257 | if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU]: 258 | m.inplace = True # pytorch 1.7.0 compatibility 259 | elif type(m) is nn.Upsample: 260 | m.recompute_scale_factor = None # torch 1.11.0 compatibility 261 | elif type(m) is Conv: 262 | m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility 263 | 264 | if len(model) == 1: 265 | return model[-1] # return model 266 | else: 267 | print('Ensemble created with %s\n' % weights) 268 | for k in ['names', 'stride']: 269 | setattr(model, k, getattr(model[-1], k)) 270 | return model # return ensemble 271 | 272 | 273 | -------------------------------------------------------------------------------- /models/quant_yolo.py: -------------------------------------------------------------------------------- 1 | import torch.quantization 2 | from .yolo import * 3 | 4 | 5 | class Model_quant_static(nn.Module): 6 | def __init__(self, cfg='yolor-csp-c.yaml', ch=3, nc=None, anchors=None): # model, input channels, number of classes 7 | super(Model_quant_static, self).__init__() 8 | self.quant = torch.quantization.QuantStub() 9 | self.dequant = torch.quantization.DeQuantStub() 10 | self.qconfig = None 11 | 12 | self.traced = False 13 | if isinstance(cfg, dict): 14 | self.yaml = cfg # model dict 15 | else: # is *.yaml 16 | import yaml # for torch hub 17 | self.yaml_file = Path(cfg).name 18 | with open(cfg) as f: 19 | self.yaml = yaml.load(f, Loader=yaml.SafeLoader) # model dict 20 | 21 | # Define model 22 | ch = self.yaml['ch'] = self.yaml.get('ch', ch) # input channels 23 | if nc and nc != self.yaml['nc']: 24 | logger.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}") 25 | self.yaml['nc'] = nc # override yaml value 26 | if anchors: 27 | logger.info(f'Overriding model.yaml anchors with anchors={anchors}') 28 | self.yaml['anchors'] = round(anchors) # override yaml value 29 | self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # model, savelist 30 | self.names = [str(i) for i in range(self.yaml['nc'])] # default names 31 | # print([x.shape for x in self.forward(torch.zeros(1, ch, 64, 64))]) 32 | 33 | # Build strides, anchors 34 | m = self.model[-1] # Detect() 35 | if isinstance(m, Detect): 36 | s = 256 # 2x min stride 37 | m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward 38 | check_anchor_order(m) 39 | m.anchors /= m.stride.view(-1, 1, 1) 40 | self.stride = m.stride 41 | self._initialize_biases() # only run once 42 | # print('Strides: %s' % m.stride.tolist()) 43 | if isinstance(m, IDetect): 44 | s = 256 # 2x min stride 45 | m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward 46 | check_anchor_order(m) 47 | m.anchors /= m.stride.view(-1, 1, 1) 48 | self.stride = m.stride 49 | self._initialize_biases() # only run once 50 | # print('Strides: %s' % m.stride.tolist()) 51 | if isinstance(m, IAuxDetect): 52 | s = 256 # 2x min stride 53 | m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))[:4]]) # forward 54 | # print(m.stride) 55 | check_anchor_order(m) 56 | m.anchors /= m.stride.view(-1, 1, 1) 57 | self.stride = m.stride 58 | self._initialize_aux_biases() # only run once 59 | # print('Strides: %s' % m.stride.tolist()) 60 | if isinstance(m, IBin): 61 | s = 256 # 2x min stride 62 | m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward 63 | check_anchor_order(m) 64 | m.anchors /= m.stride.view(-1, 1, 1) 65 | self.stride = m.stride 66 | self._initialize_biases_bin() # only run once 67 | # print('Strides: %s' % m.stride.tolist()) 68 | if isinstance(m, IKeypoint): 69 | s = 256 # 2x min stride 70 | m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward 71 | check_anchor_order(m) 72 | m.anchors /= m.stride.view(-1, 1, 1) 73 | self.stride = m.stride 74 | self._initialize_biases_kpt() # only run once 75 | # print('Strides: %s' % m.stride.tolist()) 76 | 77 | # Init weights, biases 78 | initialize_weights(self) 79 | self.info() 80 | logger.info('') 81 | 82 | def forward(self, x, augment=False, profile=False): 83 | x = self.quant(x) 84 | 85 | if augment: 86 | img_size = x.shape[-2:] # height, width 87 | s = [1, 0.83, 0.67] # scales 88 | f = [None, 3, None] # flips (2-ud, 3-lr) 89 | y = [] # outputs 90 | for si, fi in zip(s, f): 91 | xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max())) 92 | yi = self.forward_once(xi)[0] # forward 93 | # cv2.imwrite(f'img_{si}.jpg', 255 * xi[0].cpu().numpy().transpose((1, 2, 0))[:, :, ::-1]) # save 94 | yi[..., :4] /= si # de-scale 95 | if fi == 2: 96 | yi[..., 1] = img_size[0] - yi[..., 1] # de-flip ud 97 | elif fi == 3: 98 | yi[..., 0] = img_size[1] - yi[..., 0] # de-flip lr 99 | y.append(yi) 100 | return self.dequant(torch.cat(y, 1)), None # augmented inference, train 101 | else: 102 | return self.dequant(self.forward_once(x, profile)) # single-scale inference, train 103 | 104 | def forward_once(self, x, profile=False): 105 | y, dt = [], [] # outputs 106 | for m in self.model: 107 | if m.f != -1: # if not from previous layer 108 | x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layers 109 | 110 | if not hasattr(self, 'traced'): 111 | self.traced = False 112 | 113 | if self.traced: 114 | if isinstance(m, Detect) or isinstance(m, IDetect) or isinstance(m, IAuxDetect) or isinstance(m, 115 | IKeypoint): 116 | break 117 | 118 | if profile: 119 | c = isinstance(m, (Detect, IDetect, IAuxDetect, IBin)) 120 | o = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0 # FLOPS 121 | for _ in range(10): 122 | m(x.copy() if c else x) 123 | t = time_synchronized() 124 | for _ in range(10): 125 | m(x.copy() if c else x) 126 | dt.append((time_synchronized() - t) * 100) 127 | print('%10.1f%10.0f%10.1fms %-40s' % (o, m.np, dt[-1], m.type)) 128 | 129 | x = m(x) # run 130 | 131 | y.append(x if m.i in self.save else None) # save output 132 | 133 | if profile: 134 | print('%.1fms total' % sum(dt)) 135 | return x 136 | 137 | def _initialize_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency 138 | # https://arxiv.org/abs/1708.02002 section 3.3 139 | # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. 140 | m = self.model[-1] # Detect() module 141 | for mi, s in zip(m.m, m.stride): # from 142 | b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) 143 | b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) 144 | b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls 145 | mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) 146 | 147 | def _initialize_aux_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency 148 | # https://arxiv.org/abs/1708.02002 section 3.3 149 | # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. 150 | m = self.model[-1] # Detect() module 151 | for mi, mi2, s in zip(m.m, m.m2, m.stride): # from 152 | b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) 153 | b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) 154 | b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls 155 | mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) 156 | b2 = mi2.bias.view(m.na, -1) # conv.bias(255) to (3,85) 157 | b2.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) 158 | b2.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls 159 | mi2.bias = torch.nn.Parameter(b2.view(-1), requires_grad=True) 160 | 161 | def _initialize_biases_bin(self, cf=None): # initialize biases into Detect(), cf is class frequency 162 | # https://arxiv.org/abs/1708.02002 section 3.3 163 | # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. 164 | m = self.model[-1] # Bin() module 165 | bc = m.bin_count 166 | for mi, s in zip(m.m, m.stride): # from 167 | b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) 168 | old = b[:, (0, 1, 2, bc + 3)].data 169 | obj_idx = 2 * bc + 4 170 | b[:, :obj_idx].data += math.log(0.6 / (bc + 1 - 0.99)) 171 | b[:, obj_idx].data += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) 172 | b[:, (obj_idx + 1):].data += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log( 173 | cf / cf.sum()) # cls 174 | b[:, (0, 1, 2, bc + 3)].data = old 175 | mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) 176 | 177 | def _initialize_biases_kpt(self, cf=None): # initialize biases into Detect(), cf is class frequency 178 | # https://arxiv.org/abs/1708.02002 section 3.3 179 | # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. 180 | m = self.model[-1] # Detect() module 181 | for mi, s in zip(m.m, m.stride): # from 182 | b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) 183 | b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) 184 | b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls 185 | mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) 186 | 187 | def _print_biases(self): 188 | m = self.model[-1] # Detect() module 189 | for mi in m.m: # from 190 | b = mi.bias.detach().view(m.na, -1).T # conv.bias(255) to (3,85) 191 | print(('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean())) 192 | 193 | # def _print_weights(self): 194 | # for m in self.model.modules(): 195 | # if type(m) is Bottleneck: 196 | # print('%10.3g' % (m.w.detach().sigmoid() * 2)) # shortcut weights 197 | 198 | def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers 199 | print('Fusing layers... ') 200 | for m in self.model.modules(): 201 | if isinstance(m, RepConv): 202 | # print(f" fuse_repvgg_block") 203 | m.fuse_repvgg_block() 204 | elif isinstance(m, RepConv_OREPA): 205 | # print(f" switch_to_deploy") 206 | m.switch_to_deploy() 207 | elif type(m) is Conv and hasattr(m, 'bn'): 208 | m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv 209 | delattr(m, 'bn') # remove batchnorm 210 | m.forward = m.fuseforward # update forward 211 | elif isinstance(m, (IDetect, IAuxDetect)): 212 | m.fuse() 213 | m.forward = m.fuseforward 214 | self.info() 215 | return self 216 | 217 | def nms(self, mode=True): # add or remove NMS module 218 | present = type(self.model[-1]) is NMS # last layer is NMS 219 | if mode and not present: 220 | print('Adding NMS... ') 221 | m = NMS() # module 222 | m.f = -1 # from 223 | m.i = self.model[-1].i + 1 # index 224 | self.model.add_module(name='%s' % m.i, module=m) # add 225 | self.eval() 226 | elif not mode and present: 227 | print('Removing NMS... ') 228 | self.model = self.model[:-1] # remove 229 | return self 230 | 231 | def autoshape(self): # add autoShape module 232 | print('Adding autoShape... ') 233 | m = autoShape(self) # wrap model 234 | copy_attr(m, self, include=('yaml', 'nc', 'hyp', 'names', 'stride'), exclude=()) # copy attributes 235 | return m 236 | 237 | def info(self, verbose=False, img_size=640): # print model information 238 | model_info(self, verbose, img_size) 239 | 240 | 241 | def dynamic_quant(_model, dtype, qconfig_spec, save_path=None): 242 | print("Dynamic quantify model------") 243 | model_quant = torch.quantization.quantize_dynamic( 244 | model=_model, 245 | qconfig_spec=qconfig_spec, 246 | dtype=dtype 247 | ) 248 | 249 | if save_path is not None: 250 | torch.save(model_quant.state_dict(), save_path) 251 | 252 | return model_quant 253 | 254 | 255 | def static_quant(_model, data_iter, deploy_device, device, save_path=None): 256 | print("Static quantify model------") 257 | _model.eval() 258 | 259 | if deploy_device == 'arm': 260 | _model.qconfig = torch.quantization.get_default_qconfig('qnnpack') 261 | elif deploy_device == 'x86': 262 | _model.qconfig = torch.quantization.get_default_qconfig('fbgemm') 263 | else: 264 | raise ValueError(f"deploy_device only support arm or x86, but got {deploy_device}") 265 | 266 | model_prepare = torch.quantization.prepare(_model) 267 | 268 | for data, _, _, _ in data_iter: 269 | model_prepare(data.to(device).float() / 255.0) 270 | del data 271 | 272 | _model = torch.quantization.convert(model_prepare) 273 | 274 | if save_path is not None: 275 | torch.save(_model.state_dict(), save_path) 276 | 277 | return _model 278 | 279 | -------------------------------------------------------------------------------- /papers/README.md: -------------------------------------------------------------------------------- 1 | # Papers： 2 | 3 | ## 剪枝 4 | 5 | ### （1）非结构剪枝 6 | 7 | 《DepGraph: Towards Any Structural Pruning》 8 | 9 | Link: [2301.12900.pdf (arxiv.org)](https://arxiv.org/pdf/2301.12900.pdf) 10 | 11 | 摘要：提出了一种通用的全自动方法——依赖图(Dependency Graph, DepGraph)来显式地对层间的相互依赖进行建模，并对耦合参数进行综合分组。 12 | 13 | Github: [WongKinYiu/yolov7: Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors (github.com)](https://github.com/WongKinYiu/yolov7) 14 | 15 | 《Network Pruning via Transformable Architecture Search》 16 | 17 | Link: [1905.09717v5.pdf (arxiv.org)](https://arxiv.org/pdf/1905.09717v5.pdf) 18 | 19 | 摘要：对于每一层网络分化出多个同类的较小网络，通过知识蒸馏让预备网络结构学习原网络的特征图表示，选出损耗最小的网络替换原网络层。 20 | 21 | Github: [D-X-Y/AutoDL-Projects: Automated deep learning algorithms implemented in PyTorch. (github.com)](https://github.com/D-X-Y/AutoDL-Projects) 22 | 23 | ### （2）结构剪枝 24 | 25 | 《Structured Pruning for Deep Convolutional Neural Networks: A survey》 26 | 27 | Link: [2303.00566v1.pdf (arxiv.org)](https://arxiv.org/pdf/2303.00566v1.pdf) 28 | 29 | 摘要：从过滤器排序方法、正则化方法、动态执行、神经结构搜索、彩票假设和剪枝的应用等方面对目前最先进的结构化剪枝技术进行了总结和比较。（综述） 30 | 31 | Github: [he-y/Awesome-Pruning: A curated list of neural network pruning resources. (github.com)](https://github.com/he-y/Awesome-Pruning) 32 | 33 | 《Movement Pruning: Adaptive Sparsity by Fine-Tuning》 34 | 35 | Link: [2005.07683v2.pdf (arxiv.org)](https://arxiv.org/pdf/2005.07683v2.pdf) 36 | 37 | 摘要：运动剪枝，在训练过程中保留重要性高的连接，即修建在训练过程中逐渐趋于0的连接，适合用来微调预训练模型，使参数稀疏化。 38 | 39 | ## 量化 40 | 41 | 《TRAINING WITH QUANTIZATION NOISE FOR EXTREME MODEL COMPRESSION》 42 | 43 | Link: [2004.07320v3.pdf (arxiv.org)](https://arxiv.org/pdf/2004.07320v3.pdf) 44 | 45 | 摘要：通过在训练中引入随机量化、部分量化增强模型对精度损失的鲁棒性。 46 | 47 | 《LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale》 48 | 49 | Link: [2208.07339v2.pdf (arxiv.org)](https://arxiv.org/pdf/2208.07339v2.pdf) 50 | 51 | 摘要：对矩阵乘法中的每个内积使用独立的归一化常数的矢量量化，以量化大多数特征，通过对列和行规范化常数的外积进行反规范化处理来恢复矩阵乘法的输出，效果近乎无损。 -------------------------------------------------------------------------------- /prune.py: -------------------------------------------------------------------------------- 1 | import time 2 | import torch 3 | import torch_pruning as tp 4 | from models.yolo import Detect, IDetect 5 | from models.common import ImplicitA, ImplicitM 6 | 7 | 8 | class pruner: 9 | def __init__(self, model, device, opt): 10 | model.eval() 11 | example_inputs = torch.randn(1, 3, 224, 224).to(device) 12 | imp = tp.importance.MagnitudeImportance(p=2 if opt.prune_norm == 'L2' else 1) # L2 norm pruning 13 | 14 | ignored_layers = [] 15 | for m in model.modules(): 16 | if isinstance(m, (Detect, IDetect)): 17 | ignored_layers.append(m.m) 18 | unwrapped_parameters = [] 19 | for m in model.modules(): 20 | if isinstance(m, (ImplicitA, ImplicitM)): 21 | unwrapped_parameters.append((m.implicit, 1)) # pruning 1st dimension of implicit matrix 22 | 23 | iterative_steps = opt.epochs // opt.num_epochs_to_prune # progressive pruning 24 | self.pruner = tp.pruner.MagnitudePruner( 25 | model, 26 | example_inputs, 27 | importance=imp, 28 | iterative_steps=iterative_steps, 29 | ch_sparsity=opt.sparsity, 30 | # remove 50% channels, ResNet18 = {64, 128, 256, 512} => ResNet18_Half = {32, 64, 128, 256} 31 | ignored_layers=ignored_layers, 32 | unwrapped_parameters=unwrapped_parameters 33 | ) 34 | self.sparsity = opt.sparsity 35 | self.num_steps = iterative_steps 36 | self.count = 0 37 | 38 | def step(self, model, device): 39 | self.count += 1 40 | 41 | example_inputs = torch.randn(1, 3, 224, 224).to(device) 42 | base_macs, base_nparams = tp.utils.count_ops_and_params(model, example_inputs) 43 | 44 | self.pruner.step() 45 | 46 | pruned_macs, pruned_nparams = tp.utils.count_ops_and_params(model, example_inputs) 47 | print("Pruning Sparsity=%f" % (self.sparsity / self.num_steps * self.count)) 48 | print("Before Pruning: MACs=%f, #Params=%f" % (base_macs, base_nparams)) 49 | print("After Pruning: MACs=%f, #Params=%f" % (pruned_macs, pruned_nparams)) 50 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | torch~=1.13.0 2 | PyYAML~=6.0 3 | opencv-python~=4.6.0.66 4 | matplotlib~=3.5.2 5 | numpy~=1.23.1 6 | pandas~=1.4.3 7 | seaborn~=0.11.2 8 | Pillow~=9.2.0 9 | scipy~=1.8.1 10 | torchvision~=0.14.0 11 | tqdm~=4.64.0 12 | requests~=2.28.1 13 | torch_pruning~=1.1.6 -------------------------------------------------------------------------------- /test.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | import json 3 | import os 4 | from pathlib import Path 5 | from threading import Thread 6 | 7 | import numpy as np 8 | import torch 9 | import yaml 10 | from tqdm import tqdm 11 | 12 | from models.experimental import attempt_load 13 | from utils.datasets import create_dataloader 14 | from utils.general import coco80_to_coco91_class, check_dataset, check_file, check_img_size, check_requirements, \ 15 | box_iou, non_max_suppression, scale_coords, xyxy2xywh, xywh2xyxy, set_logging, increment_path, colorstr 16 | from utils.metrics import ap_per_class, ConfusionMatrix 17 | from utils.plots import plot_images, output_to_target, plot_study_txt 18 | from utils.torch_utils import select_device, time_synchronized, TracedModel 19 | 20 | 21 | def test(data, 22 | weights=None, 23 | batch_size=32, 24 | imgsz=640, 25 | conf_thres=0.001, 26 | iou_thres=0.6, # for NMS 27 | save_json=False, 28 | single_cls=False, 29 | augment=False, 30 | verbose=False, 31 | model=None, 32 | dataloader=None, 33 | save_dir=Path(''), # for saving images 34 | save_txt=False, # for auto-labelling 35 | save_hybrid=False, # for hybrid auto-labelling 36 | save_conf=False, # save auto-label confidences 37 | plots=True, 38 | wandb_logger=None, 39 | compute_loss=None, 40 | half_precision=True, 41 | trace=False, 42 | is_coco=False, 43 | v5_metric=False): 44 | # Initialize/load model and set device 45 | training = model is not None 46 | if training: # called by train.py 47 | device = next(model.parameters()).device # get model device 48 | 49 | else: # called directly 50 | set_logging() 51 | device = select_device(opt.device, batch_size=batch_size) 52 | 53 | # Directories 54 | save_dir = Path(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok)) # increment run 55 | (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True) # make dir 56 | 57 | # Load model 58 | model = attempt_load(weights, map_location=device) # load FP32 model 59 | gs = max(int(model.stride.max()), 32) # grid size (max stride) 60 | imgsz = check_img_size(imgsz, s=gs) # check img_size 61 | 62 | if trace: 63 | model = TracedModel(model, device, imgsz) 64 | 65 | # Half 66 | half = device.type != 'cpu' and half_precision # half precision only supported on CUDA 67 | if half: 68 | model.half() 69 | 70 | # Configure 71 | model.eval() 72 | if isinstance(data, str): 73 | is_coco = data.endswith('coco.yaml') 74 | with open(data) as f: 75 | data = yaml.load(f, Loader=yaml.SafeLoader) 76 | check_dataset(data) # check 77 | nc = 1 if single_cls else int(data['nc']) # number of classes 78 | iouv = torch.linspace(0.5, 0.95, 10).to(device) # iou vector for mAP@0.5:0.95 79 | niou = iouv.numel() 80 | 81 | # Logging 82 | log_imgs = 0 83 | if wandb_logger and wandb_logger.wandb: 84 | log_imgs = min(wandb_logger.log_imgs, 100) 85 | # Dataloader 86 | if not training: 87 | if device.type != 'cpu': 88 | model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters()))) # run once 89 | task = opt.task if opt.task in ('train', 'val', 'test') else 'val' # path to train/val/test images 90 | dataloader = create_dataloader(data[task], imgsz, batch_size, gs, opt, pad=0.5, rect=True, 91 | prefix=colorstr(f'{task}: '))[0] 92 | 93 | if v5_metric: 94 | print("Testing with YOLOv5 AP metric...") 95 | 96 | seen = 0 97 | confusion_matrix = ConfusionMatrix(nc=nc) 98 | names = {k: v for k, v in enumerate(model.names if hasattr(model, 'names') else model.module.names)} 99 | coco91class = coco80_to_coco91_class() 100 | s = ('%20s' + '%12s' * 6) % ('Class', 'Images', 'Labels', 'P', 'R', 'mAP@.5', 'mAP@.5:.95') 101 | p, r, f1, mp, mr, map50, map, t0, t1 = 0., 0., 0., 0., 0., 0., 0., 0., 0. 102 | loss = torch.zeros(3, device=device) 103 | jdict, stats, ap, ap_class, wandb_images = [], [], [], [], [] 104 | for batch_i, (img, targets, paths, shapes) in enumerate(tqdm(dataloader, desc=s)): 105 | img = img.to(device, non_blocking=True) 106 | img = img.half() if half else img.float() # uint8 to fp16/32 107 | img /= 255.0 # 0 - 255 to 0.0 - 1.0 108 | targets = targets.to(device) 109 | nb, _, height, width = img.shape # batch size, channels, height, width 110 | 111 | with torch.no_grad(): 112 | # Run model 113 | t = time_synchronized() 114 | out, train_out = model(img, augment=augment) # inference and training outputs 115 | t0 += time_synchronized() - t 116 | 117 | # Compute loss 118 | if compute_loss: 119 | loss += compute_loss([x.float() for x in train_out], targets)[1][:3] # box, obj, cls 120 | 121 | # Run NMS 122 | targets[:, 2:] *= torch.Tensor([width, height, width, height]).to(device) # to pixels 123 | lb = [targets[targets[:, 0] == i, 1:] for i in range(nb)] if save_hybrid else [] # for autolabelling 124 | t = time_synchronized() 125 | out = non_max_suppression(out, conf_thres=conf_thres, iou_thres=iou_thres, labels=lb, multi_label=True) 126 | t1 += time_synchronized() - t 127 | 128 | # Statistics per image 129 | for si, pred in enumerate(out): 130 | labels = targets[targets[:, 0] == si, 1:] 131 | nl = len(labels) 132 | tcls = labels[:, 0].tolist() if nl else [] # target class 133 | path = Path(paths[si]) 134 | seen += 1 135 | 136 | if len(pred) == 0: 137 | if nl: 138 | stats.append((torch.zeros(0, niou, dtype=torch.bool), torch.Tensor(), torch.Tensor(), tcls)) 139 | continue 140 | 141 | # Predictions 142 | predn = pred.clone() 143 | scale_coords(img[si].shape[1:], predn[:, :4], shapes[si][0], shapes[si][1]) # native-space pred 144 | 145 | # Append to text file 146 | if save_txt: 147 | gn = torch.tensor(shapes[si][0])[[1, 0, 1, 0]] # normalization gain whwh 148 | for *xyxy, conf, cls in predn.tolist(): 149 | xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh 150 | line = (cls, *xywh, conf) if save_conf else (cls, *xywh) # label format 151 | with open(save_dir / 'labels' / (path.stem + '.txt'), 'a') as f: 152 | f.write(('%g ' * len(line)).rstrip() % line + '\n') 153 | 154 | # W&B logging - Media Panel Plots 155 | if len(wandb_images) < log_imgs and wandb_logger.current_epoch > 0: # Check for test operation 156 | if wandb_logger.current_epoch % wandb_logger.bbox_interval == 0: 157 | box_data = [{"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]}, 158 | "class_id": int(cls), 159 | "box_caption": "%s %.3f" % (names[cls], conf), 160 | "scores": {"class_score": conf}, 161 | "domain": "pixel"} for *xyxy, conf, cls in pred.tolist()] 162 | boxes = {"predictions": {"box_data": box_data, "class_labels": names}} # inference-space 163 | wandb_images.append(wandb_logger.wandb.Image(img[si], boxes=boxes, caption=path.name)) 164 | wandb_logger.log_training_progress(predn, path, names) if wandb_logger and wandb_logger.wandb_run else None 165 | 166 | # Append to pycocotools JSON dictionary 167 | if save_json: 168 | # [{"image_id": 42, "category_id": 18, "bbox": [258.15, 41.29, 348.26, 243.78], "score": 0.236}, ... 169 | image_id = int(path.stem) if path.stem.isnumeric() else path.stem 170 | box = xyxy2xywh(predn[:, :4]) # xywh 171 | box[:, :2] -= box[:, 2:] / 2 # xy center to top-left corner 172 | for p, b in zip(pred.tolist(), box.tolist()): 173 | jdict.append({'image_id': image_id, 174 | 'category_id': coco91class[int(p[5])] if is_coco else int(p[5]), 175 | 'bbox': [round(x, 3) for x in b], 176 | 'score': round(p[4], 5)}) 177 | 178 | # Assign all predictions as incorrect 179 | correct = torch.zeros(pred.shape[0], niou, dtype=torch.bool, device=device) 180 | if nl: 181 | detected = [] # target indices 182 | tcls_tensor = labels[:, 0] 183 | 184 | # target boxes 185 | tbox = xywh2xyxy(labels[:, 1:5]) 186 | scale_coords(img[si].shape[1:], tbox, shapes[si][0], shapes[si][1]) # native-space labels 187 | if plots: 188 | confusion_matrix.process_batch(predn, torch.cat((labels[:, 0:1], tbox), 1)) 189 | 190 | # Per target class 191 | for cls in torch.unique(tcls_tensor): 192 | ti = (cls == tcls_tensor).nonzero(as_tuple=False).view(-1) # prediction indices 193 | pi = (cls == pred[:, 5]).nonzero(as_tuple=False).view(-1) # target indices 194 | 195 | # Search for detections 196 | if pi.shape[0]: 197 | # Prediction to target ious 198 | ious, i = box_iou(predn[pi, :4], tbox[ti]).max(1) # best ious, indices 199 | 200 | # Append detections 201 | detected_set = set() 202 | for j in (ious > iouv[0]).nonzero(as_tuple=False): 203 | d = ti[i[j]] # detected target 204 | if d.item() not in detected_set: 205 | detected_set.add(d.item()) 206 | detected.append(d) 207 | correct[pi[j]] = ious[j] > iouv # iou_thres is 1xn 208 | if len(detected) == nl: # all targets already located in image 209 | break 210 | 211 | # Append statistics (correct, conf, pcls, tcls) 212 | stats.append((correct.cpu(), pred[:, 4].cpu(), pred[:, 5].cpu(), tcls)) 213 | 214 | # Plot images 215 | if plots and batch_i < 3: 216 | f = save_dir / f'test_batch{batch_i}_labels.jpg' # labels 217 | Thread(target=plot_images, args=(img, targets, paths, f, names), daemon=True).start() 218 | f = save_dir / f'test_batch{batch_i}_pred.jpg' # predictions 219 | Thread(target=plot_images, args=(img, output_to_target(out), paths, f, names), daemon=True).start() 220 | 221 | # Compute statistics 222 | stats = [np.concatenate(x, 0) for x in zip(*stats)] # to numpy 223 | if len(stats) and stats[0].any(): 224 | p, r, ap, f1, ap_class = ap_per_class(*stats, plot=plots, v5_metric=v5_metric, save_dir=save_dir, names=names) 225 | ap50, ap = ap[:, 0], ap.mean(1) # AP@0.5, AP@0.5:0.95 226 | mp, mr, map50, map = p.mean(), r.mean(), ap50.mean(), ap.mean() 227 | nt = np.bincount(stats[3].astype(np.int64), minlength=nc) # number of targets per class 228 | else: 229 | nt = torch.zeros(1) 230 | 231 | # Print results 232 | pf = '%20s' + '%12i' * 2 + '%12.3g' * 4 # print format 233 | print(pf % ('all', seen, nt.sum(), mp, mr, map50, map)) 234 | 235 | # Print results per class 236 | if (verbose or (nc < 50 and not training)) and nc > 1 and len(stats): 237 | for i, c in enumerate(ap_class): 238 | print(pf % (names[c], seen, nt[c], p[i], r[i], ap50[i], ap[i])) 239 | 240 | # Print speeds 241 | t = tuple(x / seen * 1E3 for x in (t0, t1, t0 + t1)) + (imgsz, imgsz, batch_size) # tuple 242 | if not training: 243 | print('Speed: %.1f/%.1f/%.1f ms inference/NMS/total per %gx%g image at batch-size %g' % t) 244 | 245 | # Plots 246 | if plots: 247 | confusion_matrix.plot(save_dir=save_dir, names=list(names.values())) 248 | if wandb_logger and wandb_logger.wandb: 249 | val_batches = [wandb_logger.wandb.Image(str(f), caption=f.name) for f in sorted(save_dir.glob('test*.jpg'))] 250 | wandb_logger.log({"Validation": val_batches}) 251 | if wandb_images: 252 | wandb_logger.log({"Bounding Box Debugger/Images": wandb_images}) 253 | 254 | # Save JSON 255 | if save_json and len(jdict): 256 | w = Path(weights[0] if isinstance(weights, list) else weights).stem if weights is not None else '' # weights 257 | anno_json = './coco/annotations/instances_val2017.json' # annotations json 258 | pred_json = str(save_dir / f"{w}_predictions.json") # predictions json 259 | print('\nEvaluating pycocotools mAP... saving %s...' % pred_json) 260 | with open(pred_json, 'w') as f: 261 | json.dump(jdict, f) 262 | 263 | try: # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb 264 | from pycocotools.coco import COCO 265 | from pycocotools.cocoeval import COCOeval 266 | 267 | anno = COCO(anno_json) # init annotations api 268 | pred = anno.loadRes(pred_json) # init predictions api 269 | eval = COCOeval(anno, pred, 'bbox') 270 | if is_coco: 271 | eval.params.imgIds = [int(Path(x).stem) for x in dataloader.dataset.img_files] # image IDs to evaluate 272 | eval.evaluate() 273 | eval.accumulate() 274 | eval.summarize() 275 | map, map50 = eval.stats[:2] # update results (mAP@0.5:0.95, mAP@0.5) 276 | except Exception as e: 277 | print(f'pycocotools unable to run: {e}') 278 | 279 | # Return results 280 | model.float() # for training 281 | if not training: 282 | s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else '' 283 | print(f"Results saved to {save_dir}{s}") 284 | maps = np.zeros(nc) + map 285 | for i, c in enumerate(ap_class): 286 | maps[c] = ap[i] 287 | return (mp, mr, map50, map, *(loss.cpu() / len(dataloader)).tolist()), maps, t 288 | 289 | 290 | if __name__ == '__main__': 291 | parser = argparse.ArgumentParser(prog='test.py') 292 | parser.add_argument('--weights', nargs='+', type=str, default='yolov7.pt', help='model.pt path(s)') 293 | parser.add_argument('--data', type=str, default='data/coco.yaml', help='*.data path') 294 | parser.add_argument('--batch-size', type=int, default=32, help='size of each image batch') 295 | parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)') 296 | parser.add_argument('--conf-thres', type=float, default=0.001, help='object confidence threshold') 297 | parser.add_argument('--iou-thres', type=float, default=0.65, help='IOU threshold for NMS') 298 | parser.add_argument('--task', default='val', help='train, val, test, speed or study') 299 | parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') 300 | parser.add_argument('--single-cls', action='store_true', help='treat as single-class dataset') 301 | parser.add_argument('--augment', action='store_true', help='augmented inference') 302 | parser.add_argument('--verbose', action='store_true', help='report mAP by class') 303 | parser.add_argument('--save-txt', action='store_true', help='save results to *.txt') 304 | parser.add_argument('--save-hybrid', action='store_true', help='save label+prediction hybrid results to *.txt') 305 | parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels') 306 | parser.add_argument('--save-json', action='store_true', help='save a cocoapi-compatible JSON results file') 307 | parser.add_argument('--project', default='runs/test', help='save to project/name') 308 | parser.add_argument('--name', default='exp', help='save to project/name') 309 | parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment') 310 | parser.add_argument('--no-trace', action='store_true', help='don`t trace model') 311 | parser.add_argument('--v5-metric', action='store_true', help='assume maximum recall as 1.0 in AP calculation') 312 | opt = parser.parse_args() 313 | opt.save_json |= opt.data.endswith('coco.yaml') 314 | opt.data = check_file(opt.data) # check file 315 | print(opt) 316 | #check_requirements() 317 | 318 | if opt.task in ('train', 'val', 'test'): # run normally 319 | test(opt.data, 320 | opt.weights, 321 | opt.batch_size, 322 | opt.img_size, 323 | opt.conf_thres, 324 | opt.iou_thres, 325 | opt.save_json, 326 | opt.single_cls, 327 | opt.augment, 328 | opt.verbose, 329 | save_txt=opt.save_txt | opt.save_hybrid, 330 | save_hybrid=opt.save_hybrid, 331 | save_conf=opt.save_conf, 332 | trace=not opt.no_trace, 333 | v5_metric=opt.v5_metric 334 | ) 335 | 336 | elif opt.task == 'speed': # speed benchmarks 337 | for w in opt.weights: 338 | test(opt.data, w, opt.batch_size, opt.img_size, 0.25, 0.45, save_json=False, plots=False, v5_metric=opt.v5_metric) 339 | 340 | elif opt.task == 'study': # run over a range of settings and save/plot 341 | # python test.py --task study --data coco.yaml --iou 0.65 --weights yolov7.pt 342 | x = list(range(256, 1536 + 128, 128)) # x axis (image sizes) 343 | for w in opt.weights: 344 | f = f'study_{Path(opt.data).stem}_{Path(w).stem}.txt' # filename to save to 345 | y = [] # y axis 346 | for i in x: # img-size 347 | print(f'\nRunning {f} point {i}...') 348 | r, _, t = test(opt.data, w, opt.batch_size, i, opt.conf_thres, opt.iou_thres, opt.save_json, 349 | plots=False, v5_metric=opt.v5_metric) 350 | y.append(r + t) # results and times 351 | np.savetxt(f, y, fmt='%10.4g') # save 352 | os.system('zip -r study.zip study_*.txt') 353 | plot_study_txt(x=x) # plot 354 | -------------------------------------------------------------------------------- /utils/__init__.py: -------------------------------------------------------------------------------- 1 | # init -------------------------------------------------------------------------------- /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 | class MemoryEfficientSwish(nn.Module): 23 | class F(torch.autograd.Function): 24 | @staticmethod 25 | def forward(ctx, x): 26 | ctx.save_for_backward(x) 27 | return x * torch.sigmoid(x) 28 | 29 | @staticmethod 30 | def backward(ctx, grad_output): 31 | x = ctx.saved_tensors[0] 32 | sx = torch.sigmoid(x) 33 | return grad_output * (sx * (1 + x * (1 - sx))) 34 | 35 | def forward(self, x): 36 | return self.F.apply(x) 37 | 38 | 39 | # Mish https://github.com/digantamisra98/Mish -------------------------------------------------------------------------- 40 | class Mish(nn.Module): 41 | @staticmethod 42 | def forward(x): 43 | return x * F.softplus(x).tanh() 44 | 45 | 46 | class MemoryEfficientMish(nn.Module): 47 | class F(torch.autograd.Function): 48 | @staticmethod 49 | def forward(ctx, x): 50 | ctx.save_for_backward(x) 51 | return x.mul(torch.tanh(F.softplus(x))) # x * tanh(ln(1 + exp(x))) 52 | 53 | @staticmethod 54 | def backward(ctx, grad_output): 55 | x = ctx.saved_tensors[0] 56 | sx = torch.sigmoid(x) 57 | fx = F.softplus(x).tanh() 58 | return grad_output * (fx + x * sx * (1 - fx * fx)) 59 | 60 | def forward(self, x): 61 | return self.F.apply(x) 62 | 63 | 64 | # FReLU https://arxiv.org/abs/2007.11824 ------------------------------------------------------------------------------- 65 | class FReLU(nn.Module): 66 | def __init__(self, c1, k=3): # ch_in, kernel 67 | super().__init__() 68 | self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False) 69 | self.bn = nn.BatchNorm2d(c1) 70 | 71 | def forward(self, x): 72 | return torch.max(x, self.bn(self.conv(x))) 73 | -------------------------------------------------------------------------------- /utils/add_nms.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import onnx 3 | from onnx import shape_inference 4 | try: 5 | import onnx_graphsurgeon as gs 6 | except Exception as e: 7 | print('Import onnx_graphsurgeon failure: %s' % e) 8 | 9 | import logging 10 | 11 | LOGGER = logging.getLogger(__name__) 12 | 13 | class RegisterNMS(object): 14 | def __init__( 15 | self, 16 | onnx_model_path: str, 17 | precision: str = "fp32", 18 | ): 19 | 20 | self.graph = gs.import_onnx(onnx.load(onnx_model_path)) 21 | assert self.graph 22 | LOGGER.info("ONNX graph created successfully") 23 | # Fold constants via ONNX-GS that PyTorch2ONNX may have missed 24 | self.graph.fold_constants() 25 | self.precision = precision 26 | self.batch_size = 1 27 | def infer(self): 28 | """ 29 | Sanitize the graph by cleaning any unconnected nodes, do a topological resort, 30 | and fold constant inputs values. When possible, run shape inference on the 31 | ONNX graph to determine tensor shapes. 32 | """ 33 | for _ in range(3): 34 | count_before = len(self.graph.nodes) 35 | 36 | self.graph.cleanup().toposort() 37 | try: 38 | for node in self.graph.nodes: 39 | for o in node.outputs: 40 | o.shape = None 41 | model = gs.export_onnx(self.graph) 42 | model = shape_inference.infer_shapes(model) 43 | self.graph = gs.import_onnx(model) 44 | except Exception as e: 45 | LOGGER.info(f"Shape inference could not be performed at this time:\n{e}") 46 | try: 47 | self.graph.fold_constants(fold_shapes=True) 48 | except TypeError as e: 49 | LOGGER.error( 50 | "This version of ONNX GraphSurgeon does not support folding shapes, " 51 | f"please upgrade your onnx_graphsurgeon module. Error:\n{e}" 52 | ) 53 | raise 54 | 55 | count_after = len(self.graph.nodes) 56 | if count_before == count_after: 57 | # No new folding occurred in this iteration, so we can stop for now. 58 | break 59 | 60 | def save(self, output_path): 61 | """ 62 | Save the ONNX model to the given location. 63 | Args: 64 | output_path: Path pointing to the location where to write 65 | out the updated ONNX model. 66 | """ 67 | self.graph.cleanup().toposort() 68 | model = gs.export_onnx(self.graph) 69 | onnx.save(model, output_path) 70 | LOGGER.info(f"Saved ONNX model to {output_path}") 71 | 72 | def register_nms( 73 | self, 74 | *, 75 | score_thresh: float = 0.25, 76 | nms_thresh: float = 0.45, 77 | detections_per_img: int = 100, 78 | ): 79 | """ 80 | Register the ``EfficientNMS_TRT`` plugin node. 81 | NMS expects these shapes for its input tensors: 82 | - box_net: [batch_size, number_boxes, 4] 83 | - class_net: [batch_size, number_boxes, number_labels] 84 | Args: 85 | score_thresh (float): The scalar threshold for score (low scoring boxes are removed). 86 | nms_thresh (float): The scalar threshold for IOU (new boxes that have high IOU 87 | overlap with previously selected boxes are removed). 88 | detections_per_img (int): Number of best detections to keep after NMS. 89 | """ 90 | 91 | self.infer() 92 | # Find the concat node at the end of the network 93 | op_inputs = self.graph.outputs 94 | op = "EfficientNMS_TRT" 95 | attrs = { 96 | "plugin_version": "1", 97 | "background_class": -1, # no background class 98 | "max_output_boxes": detections_per_img, 99 | "score_threshold": score_thresh, 100 | "iou_threshold": nms_thresh, 101 | "score_activation": False, 102 | "box_coding": 0, 103 | } 104 | 105 | if self.precision == "fp32": 106 | dtype_output = np.float32 107 | elif self.precision == "fp16": 108 | dtype_output = np.float16 109 | else: 110 | raise NotImplementedError(f"Currently not supports precision: {self.precision}") 111 | 112 | # NMS Outputs 113 | output_num_detections = gs.Variable( 114 | name="num_dets", 115 | dtype=np.int32, 116 | shape=[self.batch_size, 1], 117 | ) # A scalar indicating the number of valid detections per batch image. 118 | output_boxes = gs.Variable( 119 | name="det_boxes", 120 | dtype=dtype_output, 121 | shape=[self.batch_size, detections_per_img, 4], 122 | ) 123 | output_scores = gs.Variable( 124 | name="det_scores", 125 | dtype=dtype_output, 126 | shape=[self.batch_size, detections_per_img], 127 | ) 128 | output_labels = gs.Variable( 129 | name="det_classes", 130 | dtype=np.int32, 131 | shape=[self.batch_size, detections_per_img], 132 | ) 133 | 134 | op_outputs = [output_num_detections, output_boxes, output_scores, output_labels] 135 | 136 | # Create the NMS Plugin node with the selected inputs. The outputs of the node will also 137 | # become the final outputs of the graph. 138 | self.graph.layer(op=op, name="batched_nms", inputs=op_inputs, outputs=op_outputs, attrs=attrs) 139 | LOGGER.info(f"Created NMS plugin '{op}' with attributes: {attrs}") 140 | 141 | self.graph.outputs = op_outputs 142 | 143 | self.infer() 144 | 145 | def save(self, output_path): 146 | """ 147 | Save the ONNX model to the given location. 148 | Args: 149 | output_path: Path pointing to the location where to write 150 | out the updated ONNX model. 151 | """ 152 | self.graph.cleanup().toposort() 153 | model = gs.export_onnx(self.graph) 154 | onnx.save(model, output_path) 155 | LOGGER.info(f"Saved ONNX model to {output_path}") 156 | -------------------------------------------------------------------------------- /utils/autoanchor.py: -------------------------------------------------------------------------------- 1 | # Auto-anchor utils 2 | 3 | import numpy as np 4 | import torch 5 | import yaml 6 | from scipy.cluster.vq import kmeans 7 | from tqdm import tqdm 8 | 9 | from utils.general import colorstr 10 | 11 | 12 | def check_anchor_order(m): 13 | # Check anchor order against stride order for YOLO Detect() module m, and correct if necessary 14 | a = m.anchor_grid.prod(-1).view(-1) # anchor area 15 | da = a[-1] - a[0] # delta a 16 | ds = m.stride[-1] - m.stride[0] # delta s 17 | if da.sign() != ds.sign(): # same order 18 | print('Reversing anchor order') 19 | m.anchors[:] = m.anchors.flip(0) 20 | m.anchor_grid[:] = m.anchor_grid.flip(0) 21 | 22 | 23 | def check_anchors(dataset, model, thr=4.0, imgsz=640): 24 | # Check anchor fit to data, recompute if necessary 25 | prefix = colorstr('autoanchor: ') 26 | print(f'\n{prefix}Analyzing anchors... ', end='') 27 | m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1] # Detect() 28 | shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True) 29 | scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1)) # augment scale 30 | wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float() # wh 31 | 32 | def metric(k): # compute metric 33 | r = wh[:, None] / k[None] 34 | x = torch.min(r, 1. / r).min(2)[0] # ratio metric 35 | best = x.max(1)[0] # best_x 36 | aat = (x > 1. / thr).float().sum(1).mean() # anchors above threshold 37 | bpr = (best > 1. / thr).float().mean() # best possible recall 38 | return bpr, aat 39 | 40 | anchors = m.anchor_grid.clone().cpu().view(-1, 2) # current anchors 41 | bpr, aat = metric(anchors) 42 | print(f'anchors/target = {aat:.2f}, Best Possible Recall (BPR) = {bpr:.4f}', end='') 43 | if bpr < 0.98: # threshold to recompute 44 | print('. Attempting to improve anchors, please wait...') 45 | na = m.anchor_grid.numel() // 2 # number of anchors 46 | try: 47 | anchors = kmean_anchors(dataset, n=na, img_size=imgsz, thr=thr, gen=1000, verbose=False) 48 | except Exception as e: 49 | print(f'{prefix}ERROR: {e}') 50 | new_bpr = metric(anchors)[0] 51 | if new_bpr > bpr: # replace anchors 52 | anchors = torch.tensor(anchors, device=m.anchors.device).type_as(m.anchors) 53 | m.anchor_grid[:] = anchors.clone().view_as(m.anchor_grid) # for inference 54 | check_anchor_order(m) 55 | m.anchors[:] = anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1) # loss 56 | print(f'{prefix}New anchors saved to model. Update model *.yaml to use these anchors in the future.') 57 | else: 58 | print(f'{prefix}Original anchors better than new anchors. Proceeding with original anchors.') 59 | print('') # newline 60 | 61 | 62 | def kmean_anchors(path='./data/coco.yaml', n=9, img_size=640, thr=4.0, gen=1000, verbose=True): 63 | """ Creates kmeans-evolved anchors from training dataset 64 | 65 | Arguments: 66 | path: path to dataset *.yaml, or a loaded dataset 67 | n: number of anchors 68 | img_size: image size used for training 69 | thr: anchor-label wh ratio threshold hyperparameter hyp['anchor_t'] used for training, default=4.0 70 | gen: generations to evolve anchors using genetic algorithm 71 | verbose: print all results 72 | 73 | Return: 74 | k: kmeans evolved anchors 75 | 76 | Usage: 77 | from utils.autoanchor import *; _ = kmean_anchors() 78 | """ 79 | thr = 1. / thr 80 | prefix = colorstr('autoanchor: ') 81 | 82 | def metric(k, wh): # compute metrics 83 | r = wh[:, None] / k[None] 84 | x = torch.min(r, 1. / r).min(2)[0] # ratio metric 85 | # x = wh_iou(wh, torch.tensor(k)) # iou metric 86 | return x, x.max(1)[0] # x, best_x 87 | 88 | def anchor_fitness(k): # mutation fitness 89 | _, best = metric(torch.tensor(k, dtype=torch.float32), wh) 90 | return (best * (best > thr).float()).mean() # fitness 91 | 92 | def print_results(k): 93 | k = k[np.argsort(k.prod(1))] # sort small to large 94 | x, best = metric(k, wh0) 95 | bpr, aat = (best > thr).float().mean(), (x > thr).float().mean() * n # best possible recall, anch > thr 96 | print(f'{prefix}thr={thr:.2f}: {bpr:.4f} best possible recall, {aat:.2f} anchors past thr') 97 | print(f'{prefix}n={n}, img_size={img_size}, metric_all={x.mean():.3f}/{best.mean():.3f}-mean/best, ' 98 | f'past_thr={x[x > thr].mean():.3f}-mean: ', end='') 99 | for i, x in enumerate(k): 100 | print('%i,%i' % (round(x[0]), round(x[1])), end=', ' if i < len(k) - 1 else '\n') # use in *.cfg 101 | return k 102 | 103 | if isinstance(path, str): # *.yaml file 104 | with open(path) as f: 105 | data_dict = yaml.load(f, Loader=yaml.SafeLoader) # model dict 106 | from utils.datasets import LoadImagesAndLabels 107 | dataset = LoadImagesAndLabels(data_dict['train'], augment=True, rect=True) 108 | else: 109 | dataset = path # dataset 110 | 111 | # Get label wh 112 | shapes = img_size * dataset.shapes / dataset.shapes.max(1, keepdims=True) 113 | wh0 = np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)]) # wh 114 | 115 | # Filter 116 | i = (wh0 < 3.0).any(1).sum() 117 | if i: 118 | print(f'{prefix}WARNING: Extremely small objects found. {i} of {len(wh0)} labels are < 3 pixels in size.') 119 | wh = wh0[(wh0 >= 2.0).any(1)] # filter > 2 pixels 120 | # wh = wh * (np.random.rand(wh.shape[0], 1) * 0.9 + 0.1) # multiply by random scale 0-1 121 | 122 | # Kmeans calculation 123 | print(f'{prefix}Running kmeans for {n} anchors on {len(wh)} points...') 124 | s = wh.std(0) # sigmas for whitening 125 | k, dist = kmeans(wh / s, n, iter=30) # points, mean distance 126 | assert len(k) == n, print(f'{prefix}ERROR: scipy.cluster.vq.kmeans requested {n} points but returned only {len(k)}') 127 | k *= s 128 | wh = torch.tensor(wh, dtype=torch.float32) # filtered 129 | wh0 = torch.tensor(wh0, dtype=torch.float32) # unfiltered 130 | k = print_results(k) 131 | 132 | # Plot 133 | # k, d = [None] * 20, [None] * 20 134 | # for i in tqdm(range(1, 21)): 135 | # k[i-1], d[i-1] = kmeans(wh / s, i) # points, mean distance 136 | # fig, ax = plt.subplots(1, 2, figsize=(14, 7), tight_layout=True) 137 | # ax = ax.ravel() 138 | # ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.') 139 | # fig, ax = plt.subplots(1, 2, figsize=(14, 7)) # plot wh 140 | # ax[0].hist(wh[wh[:, 0]<100, 0],400) 141 | # ax[1].hist(wh[wh[:, 1]<100, 1],400) 142 | # fig.savefig('wh.png', dpi=200) 143 | 144 | # Evolve 145 | npr = np.random 146 | f, sh, mp, s = anchor_fitness(k), k.shape, 0.9, 0.1 # fitness, generations, mutation prob, sigma 147 | pbar = tqdm(range(gen), desc=f'{prefix}Evolving anchors with Genetic Algorithm:') # progress bar 148 | for _ in pbar: 149 | v = np.ones(sh) 150 | while (v == 1).all(): # mutate until a change occurs (prevent duplicates) 151 | v = ((npr.random(sh) < mp) * npr.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0) 152 | kg = (k.copy() * v).clip(min=2.0) 153 | fg = anchor_fitness(kg) 154 | if fg > f: 155 | f, k = fg, kg.copy() 156 | pbar.desc = f'{prefix}Evolving anchors with Genetic Algorithm: fitness = {f:.4f}' 157 | if verbose: 158 | print_results(k) 159 | 160 | return print_results(k) 161 | -------------------------------------------------------------------------------- /utils/aws/__init__.py: -------------------------------------------------------------------------------- 1 | #init -------------------------------------------------------------------------------- /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 | -------------------------------------------------------------------------------- /utils/aws/resume.py: -------------------------------------------------------------------------------- 1 | # Resume all interrupted trainings in yolor/ 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.load(f, Loader=yaml.SafeLoader) 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 | -------------------------------------------------------------------------------- /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 yolor ]; then 9 | echo "Running first-time script." # install dependencies, download COCO, pull Docker 10 | git clone -b main https://github.com/WongKinYiu/yolov7 && sudo chmod -R 777 yolov7 11 | cd yolov7 12 | bash data/scripts/get_coco.sh && echo "Data done." & 13 | sudo docker pull nvcr.io/nvidia/pytorch:21.08-py3 && 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 | -------------------------------------------------------------------------------- /utils/general.py: -------------------------------------------------------------------------------- 1 | # YOLOR general utils 2 | 3 | import glob 4 | import logging 5 | import math 6 | import os 7 | import platform 8 | import random 9 | import re 10 | import subprocess 11 | import time 12 | from pathlib import Path 13 | 14 | import cv2 15 | import numpy as np 16 | import pandas as pd 17 | import torch 18 | import torchvision 19 | import yaml 20 | 21 | from utils.google_utils import gsutil_getsize 22 | from utils.metrics import fitness 23 | from utils.torch_utils import init_torch_seeds 24 | 25 | # Settings 26 | torch.set_printoptions(linewidth=320, precision=5, profile='long') 27 | np.set_printoptions(linewidth=320, formatter={'float_kind': '{:11.5g}'.format}) # format short g, %precision=5 28 | pd.options.display.max_columns = 10 29 | cv2.setNumThreads(0) # prevent OpenCV from multithreading (incompatible with PyTorch DataLoader) 30 | os.environ['NUMEXPR_MAX_THREADS'] = str(min(os.cpu_count(), 8)) # NumExpr max threads 31 | 32 | 33 | def set_logging(rank=-1): 34 | logging.basicConfig( 35 | format="%(message)s", 36 | level=logging.INFO if rank in [-1, 0] else logging.WARN) 37 | 38 | 39 | def init_seeds(seed=0): 40 | # Initialize random number generator (RNG) seeds 41 | random.seed(seed) 42 | np.random.seed(seed) 43 | init_torch_seeds(seed) 44 | 45 | 46 | def get_latest_run(search_dir='.'): 47 | # Return path to most recent 'last.pt' in /runs (i.e. to --resume from) 48 | last_list = glob.glob(f'{search_dir}/**/last*.pt', recursive=True) 49 | return max(last_list, key=os.path.getctime) if last_list else '' 50 | 51 | 52 | def isdocker(): 53 | # Is environment a Docker container 54 | return Path('/workspace').exists() # or Path('/.dockerenv').exists() 55 | 56 | 57 | def emojis(str=''): 58 | # Return platform-dependent emoji-safe version of string 59 | return str.encode().decode('ascii', 'ignore') if platform.system() == 'Windows' else str 60 | 61 | 62 | def check_online(): 63 | # Check internet connectivity 64 | import socket 65 | try: 66 | socket.create_connection(("1.1.1.1", 443), 5) # check host accesability 67 | return True 68 | except OSError: 69 | return False 70 | 71 | 72 | def check_git_status(): 73 | # Recommend 'git pull' if code is out of date 74 | print(colorstr('github: '), end='') 75 | try: 76 | assert Path('.git').exists(), 'skipping check (not a git repository)' 77 | assert not isdocker(), 'skipping check (Docker image)' 78 | assert check_online(), 'skipping check (offline)' 79 | 80 | cmd = 'git fetch && git config --get remote.origin.url' 81 | url = subprocess.check_output(cmd, shell=True).decode().strip().rstrip('.git') # github repo url 82 | branch = subprocess.check_output('git rev-parse --abbrev-ref HEAD', shell=True).decode().strip() # checked out 83 | n = int(subprocess.check_output(f'git rev-list {branch}..origin/master --count', shell=True)) # commits behind 84 | if n > 0: 85 | s = f"⚠️ WARNING: code is out of date by {n} commit{'s' * (n > 1)}. " \ 86 | f"Use 'git pull' to update or 'git clone {url}' to download latest." 87 | else: 88 | s = f'up to date with {url} ✅' 89 | print(emojis(s)) # emoji-safe 90 | except Exception as e: 91 | print(e) 92 | 93 | 94 | def check_requirements(requirements='requirements.txt', exclude=()): 95 | # Check installed dependencies meet requirements (pass *.txt file or list of packages) 96 | import pkg_resources as pkg 97 | prefix = colorstr('red', 'bold', 'requirements:') 98 | if isinstance(requirements, (str, Path)): # requirements.txt file 99 | file = Path(requirements) 100 | if not file.exists(): 101 | print(f"{prefix} {file.resolve()} not found, check failed.") 102 | return 103 | requirements = [f'{x.name}{x.specifier}' for x in pkg.parse_requirements(file.open()) if x.name not in exclude] 104 | else: # list or tuple of packages 105 | requirements = [x for x in requirements if x not in exclude] 106 | 107 | n = 0 # number of packages updates 108 | for r in requirements: 109 | try: 110 | pkg.require(r) 111 | except Exception as e: # DistributionNotFound or VersionConflict if requirements not met 112 | n += 1 113 | print(f"{prefix} {e.req} not found and is required by YOLOR, attempting auto-update...") 114 | print(subprocess.check_output(f"pip install '{e.req}'", shell=True).decode()) 115 | 116 | if n: # if packages updated 117 | source = file.resolve() if 'file' in locals() else requirements 118 | s = f"{prefix} {n} package{'s' * (n > 1)} updated per {source}\n" \ 119 | f"{prefix} ⚠️ {colorstr('bold', 'Restart runtime or rerun command for updates to take effect')}\n" 120 | print(emojis(s)) # emoji-safe 121 | 122 | 123 | def check_img_size(img_size, s=32): 124 | # Verify img_size is a multiple of stride s 125 | new_size = make_divisible(img_size, int(s)) # ceil gs-multiple 126 | if new_size != img_size: 127 | print('WARNING: --img-size %g must be multiple of max stride %g, updating to %g' % (img_size, s, new_size)) 128 | return new_size 129 | 130 | 131 | def check_imshow(): 132 | # Check if environment supports image displays 133 | try: 134 | assert not isdocker(), 'cv2.imshow() is disabled in Docker environments' 135 | cv2.imshow('test', np.zeros((1, 1, 3))) 136 | cv2.waitKey(1) 137 | cv2.destroyAllWindows() 138 | cv2.waitKey(1) 139 | return True 140 | except Exception as e: 141 | print(f'WARNING: Environment does not support cv2.imshow() or PIL Image.show() image displays\n{e}') 142 | return False 143 | 144 | 145 | def check_file(file): 146 | # Search for file if not found 147 | if Path(file).is_file() or file == '': 148 | return file 149 | else: 150 | files = glob.glob('./**/' + file, recursive=True) # find file 151 | assert len(files), f'File Not Found: {file}' # assert file was found 152 | assert len(files) == 1, f"Multiple files match '{file}', specify exact path: {files}" # assert unique 153 | return files[0] # return file 154 | 155 | 156 | def check_dataset(dict): 157 | # Download dataset if not found locally 158 | val, s = dict.get('val'), dict.get('download') 159 | if val and len(val): 160 | val = [Path(x).resolve() for x in (val if isinstance(val, list) else [val])] # val path 161 | if not all(x.exists() for x in val): 162 | print('\nWARNING: Dataset not found, nonexistent paths: %s' % [str(x) for x in val if not x.exists()]) 163 | if s and len(s): # download script 164 | print('Downloading %s ...' % s) 165 | if s.startswith('http') and s.endswith('.zip'): # URL 166 | f = Path(s).name # filename 167 | torch.hub.download_url_to_file(s, f) 168 | r = os.system('unzip -q %s -d ../ && rm %s' % (f, f)) # unzip 169 | else: # bash script 170 | r = os.system(s) 171 | print('Dataset autodownload %s\n' % ('success' if r == 0 else 'failure')) # analyze return value 172 | else: 173 | raise Exception('Dataset not found.') 174 | 175 | 176 | def make_divisible(x, divisor): 177 | # Returns x evenly divisible by divisor 178 | return math.ceil(x / divisor) * divisor 179 | 180 | 181 | def clean_str(s): 182 | # Cleans a string by replacing special characters with underscore _ 183 | return re.sub(pattern="[|@#!¡·$€%&()=?¿^*;:,¨´><+]", repl="_", string=s) 184 | 185 | 186 | def one_cycle(y1=0.0, y2=1.0, steps=100): 187 | # lambda function for sinusoidal ramp from y1 to y2 188 | return lambda x: ((1 - math.cos(x * math.pi / steps)) / 2) * (y2 - y1) + y1 189 | 190 | 191 | def colorstr(*input): 192 | # Colors a string https://en.wikipedia.org/wiki/ANSI_escape_code, i.e. colorstr('blue', 'hello world') 193 | *args, string = input if len(input) > 1 else ('blue', 'bold', input[0]) # color arguments, string 194 | colors = {'black': '\033[30m', # basic colors 195 | 'red': '\033[31m', 196 | 'green': '\033[32m', 197 | 'yellow': '\033[33m', 198 | 'blue': '\033[34m', 199 | 'magenta': '\033[35m', 200 | 'cyan': '\033[36m', 201 | 'white': '\033[37m', 202 | 'bright_black': '\033[90m', # bright colors 203 | 'bright_red': '\033[91m', 204 | 'bright_green': '\033[92m', 205 | 'bright_yellow': '\033[93m', 206 | 'bright_blue': '\033[94m', 207 | 'bright_magenta': '\033[95m', 208 | 'bright_cyan': '\033[96m', 209 | 'bright_white': '\033[97m', 210 | 'end': '\033[0m', # misc 211 | 'bold': '\033[1m', 212 | 'underline': '\033[4m'} 213 | return ''.join(colors[x] for x in args) + f'{string}' + colors['end'] 214 | 215 | 216 | def labels_to_class_weights(labels, nc=80): 217 | # Get class weights (inverse frequency) from training labels 218 | if labels[0] is None: # no labels loaded 219 | return torch.Tensor() 220 | 221 | labels = np.concatenate(labels, 0) # labels.shape = (866643, 5) for COCO 222 | classes = labels[:, 0].astype(np.int32) # labels = [class xywh] 223 | weights = np.bincount(classes, minlength=nc) # occurrences per class 224 | 225 | # Prepend gridpoint count (for uCE training) 226 | # gpi = ((320 / 32 * np.array([1, 2, 4])) ** 2 * 3).sum() # gridpoints per image 227 | # weights = np.hstack([gpi * len(labels) - weights.sum() * 9, weights * 9]) ** 0.5 # prepend gridpoints to start 228 | 229 | weights[weights == 0] = 1 # replace empty bins with 1 230 | weights = 1 / weights # number of targets per class 231 | weights /= weights.sum() # normalize 232 | return torch.from_numpy(weights) 233 | 234 | 235 | def labels_to_image_weights(labels, nc=80, class_weights=np.ones(80)): 236 | # Produces image weights based on class_weights and image contents 237 | class_counts = np.array([np.bincount(x[:, 0].astype(np.int32), minlength=nc) for x in labels]) 238 | image_weights = (class_weights.reshape(1, nc) * class_counts).sum(1) 239 | # index = random.choices(range(n), weights=image_weights, k=1) # weight image sample 240 | return image_weights 241 | 242 | 243 | def coco80_to_coco91_class(): # converts 80-index (val2014) to 91-index (paper) 244 | # https://tech.amikelive.com/node-718/what-object-categories-labels-are-in-coco-dataset/ 245 | # a = np.loadtxt('data/coco.names', dtype='str', delimiter='\n') 246 | # b = np.loadtxt('data/coco_paper.names', dtype='str', delimiter='\n') 247 | # x1 = [list(a[i] == b).index(True) + 1 for i in range(80)] # darknet to coco 248 | # x2 = [list(b[i] == a).index(True) if any(b[i] == a) else None for i in range(91)] # coco to darknet 249 | x = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 31, 32, 33, 34, 250 | 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 251 | 64, 65, 67, 70, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 84, 85, 86, 87, 88, 89, 90] 252 | return x 253 | 254 | 255 | def xyxy2xywh(x): 256 | # Convert nx4 boxes from [x1, y1, x2, y2] to [x, y, w, h] where xy1=top-left, xy2=bottom-right 257 | y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) 258 | y[:, 0] = (x[:, 0] + x[:, 2]) / 2 # x center 259 | y[:, 1] = (x[:, 1] + x[:, 3]) / 2 # y center 260 | y[:, 2] = x[:, 2] - x[:, 0] # width 261 | y[:, 3] = x[:, 3] - x[:, 1] # height 262 | return y 263 | 264 | 265 | def xywh2xyxy(x): 266 | # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right 267 | y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) 268 | y[:, 0] = x[:, 0] - x[:, 2] / 2 # top left x 269 | y[:, 1] = x[:, 1] - x[:, 3] / 2 # top left y 270 | y[:, 2] = x[:, 0] + x[:, 2] / 2 # bottom right x 271 | y[:, 3] = x[:, 1] + x[:, 3] / 2 # bottom right y 272 | return y 273 | 274 | 275 | def xywhn2xyxy(x, w=640, h=640, padw=0, padh=0): 276 | # Convert nx4 boxes from [x, y, w, h] normalized to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right 277 | y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) 278 | y[:, 0] = w * (x[:, 0] - x[:, 2] / 2) + padw # top left x 279 | y[:, 1] = h * (x[:, 1] - x[:, 3] / 2) + padh # top left y 280 | y[:, 2] = w * (x[:, 0] + x[:, 2] / 2) + padw # bottom right x 281 | y[:, 3] = h * (x[:, 1] + x[:, 3] / 2) + padh # bottom right y 282 | return y 283 | 284 | 285 | def xyn2xy(x, w=640, h=640, padw=0, padh=0): 286 | # Convert normalized segments into pixel segments, shape (n,2) 287 | y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) 288 | y[:, 0] = w * x[:, 0] + padw # top left x 289 | y[:, 1] = h * x[:, 1] + padh # top left y 290 | return y 291 | 292 | 293 | def segment2box(segment, width=640, height=640): 294 | # Convert 1 segment label to 1 box label, applying inside-image constraint, i.e. (xy1, xy2, ...) to (xyxy) 295 | x, y = segment.T # segment xy 296 | inside = (x >= 0) & (y >= 0) & (x <= width) & (y <= height) 297 | x, y, = x[inside], y[inside] 298 | return np.array([x.min(), y.min(), x.max(), y.max()]) if any(x) else np.zeros((1, 4)) # xyxy 299 | 300 | 301 | def segments2boxes(segments): 302 | # Convert segment labels to box labels, i.e. (cls, xy1, xy2, ...) to (cls, xywh) 303 | boxes = [] 304 | for s in segments: 305 | x, y = s.T # segment xy 306 | boxes.append([x.min(), y.min(), x.max(), y.max()]) # cls, xyxy 307 | return xyxy2xywh(np.array(boxes)) # cls, xywh 308 | 309 | 310 | def resample_segments(segments, n=1000): 311 | # Up-sample an (n,2) segment 312 | for i, s in enumerate(segments): 313 | s = np.concatenate((s, s[0:1, :]), axis=0) 314 | x = np.linspace(0, len(s) - 1, n) 315 | xp = np.arange(len(s)) 316 | segments[i] = np.concatenate([np.interp(x, xp, s[:, i]) for i in range(2)]).reshape(2, -1).T # segment xy 317 | return segments 318 | 319 | 320 | def scale_coords(img1_shape, coords, img0_shape, ratio_pad=None): 321 | # Rescale coords (xyxy) from img1_shape to img0_shape 322 | if ratio_pad is None: # calculate from img0_shape 323 | gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1]) # gain = old / new 324 | pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2 # wh padding 325 | else: 326 | gain = ratio_pad[0][0] 327 | pad = ratio_pad[1] 328 | 329 | coords[:, [0, 2]] -= pad[0] # x padding 330 | coords[:, [1, 3]] -= pad[1] # y padding 331 | coords[:, :4] /= gain 332 | clip_coords(coords, img0_shape) 333 | return coords 334 | 335 | 336 | def clip_coords(boxes, img_shape): 337 | # Clip bounding xyxy bounding boxes to image shape (height, width) 338 | boxes[:, 0].clamp_(0, img_shape[1]) # x1 339 | boxes[:, 1].clamp_(0, img_shape[0]) # y1 340 | boxes[:, 2].clamp_(0, img_shape[1]) # x2 341 | boxes[:, 3].clamp_(0, img_shape[0]) # y2 342 | 343 | 344 | def bbox_iou(box1, box2, x1y1x2y2=True, GIoU=False, DIoU=False, CIoU=False, eps=1e-7): 345 | # Returns the IoU of box1 to box2. box1 is 4, box2 is nx4 346 | box2 = box2.T 347 | 348 | # Get the coordinates of bounding boxes 349 | if x1y1x2y2: # x1, y1, x2, y2 = box1 350 | b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3] 351 | b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3] 352 | else: # transform from xywh to xyxy 353 | b1_x1, b1_x2 = box1[0] - box1[2] / 2, box1[0] + box1[2] / 2 354 | b1_y1, b1_y2 = box1[1] - box1[3] / 2, box1[1] + box1[3] / 2 355 | b2_x1, b2_x2 = box2[0] - box2[2] / 2, box2[0] + box2[2] / 2 356 | b2_y1, b2_y2 = box2[1] - box2[3] / 2, box2[1] + box2[3] / 2 357 | 358 | # Intersection area 359 | inter = (torch.min(b1_x2, b2_x2) - torch.max(b1_x1, b2_x1)).clamp(0) * \ 360 | (torch.min(b1_y2, b2_y2) - torch.max(b1_y1, b2_y1)).clamp(0) 361 | 362 | # Union Area 363 | w1, h1 = b1_x2 - b1_x1, b1_y2 - b1_y1 + eps 364 | w2, h2 = b2_x2 - b2_x1, b2_y2 - b2_y1 + eps 365 | union = w1 * h1 + w2 * h2 - inter + eps 366 | 367 | iou = inter / union 368 | 369 | if GIoU or DIoU or CIoU: 370 | cw = torch.max(b1_x2, b2_x2) - torch.min(b1_x1, b2_x1) # convex (smallest enclosing box) width 371 | ch = torch.max(b1_y2, b2_y2) - torch.min(b1_y1, b2_y1) # convex height 372 | if CIoU or DIoU: # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1 373 | c2 = cw ** 2 + ch ** 2 + eps # convex diagonal squared 374 | rho2 = ((b2_x1 + b2_x2 - b1_x1 - b1_x2) ** 2 + 375 | (b2_y1 + b2_y2 - b1_y1 - b1_y2) ** 2) / 4 # center distance squared 376 | if DIoU: 377 | return iou - rho2 / c2 # DIoU 378 | elif CIoU: # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47 379 | v = (4 / math.pi ** 2) * torch.pow(torch.atan(w2 / (h2 + eps)) - torch.atan(w1 / (h1 + eps)), 2) 380 | with torch.no_grad(): 381 | alpha = v / (v - iou + (1 + eps)) 382 | return iou - (rho2 / c2 + v * alpha) # CIoU 383 | else: # GIoU https://arxiv.org/pdf/1902.09630.pdf 384 | c_area = cw * ch + eps # convex area 385 | return iou - (c_area - union) / c_area # GIoU 386 | else: 387 | return iou # IoU 388 | 389 | 390 | 391 | 392 | def bbox_alpha_iou(box1, box2, x1y1x2y2=False, GIoU=False, DIoU=False, CIoU=False, alpha=2, eps=1e-9): 393 | # Returns tsqrt_he IoU of box1 to box2. box1 is 4, box2 is nx4 394 | box2 = box2.T 395 | 396 | # Get the coordinates of bounding boxes 397 | if x1y1x2y2: # x1, y1, x2, y2 = box1 398 | b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3] 399 | b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3] 400 | else: # transform from xywh to xyxy 401 | b1_x1, b1_x2 = box1[0] - box1[2] / 2, box1[0] + box1[2] / 2 402 | b1_y1, b1_y2 = box1[1] - box1[3] / 2, box1[1] + box1[3] / 2 403 | b2_x1, b2_x2 = box2[0] - box2[2] / 2, box2[0] + box2[2] / 2 404 | b2_y1, b2_y2 = box2[1] - box2[3] / 2, box2[1] + box2[3] / 2 405 | 406 | # Intersection area 407 | inter = (torch.min(b1_x2, b2_x2) - torch.max(b1_x1, b2_x1)).clamp(0) * \ 408 | (torch.min(b1_y2, b2_y2) - torch.max(b1_y1, b2_y1)).clamp(0) 409 | 410 | # Union Area 411 | w1, h1 = b1_x2 - b1_x1, b1_y2 - b1_y1 + eps 412 | w2, h2 = b2_x2 - b2_x1, b2_y2 - b2_y1 + eps 413 | union = w1 * h1 + w2 * h2 - inter + eps 414 | 415 | # change iou into pow(iou+eps) 416 | # iou = inter / union 417 | iou = torch.pow(inter/union + eps, alpha) 418 | # beta = 2 * alpha 419 | if GIoU or DIoU or CIoU: 420 | cw = torch.max(b1_x2, b2_x2) - torch.min(b1_x1, b2_x1) # convex (smallest enclosing box) width 421 | ch = torch.max(b1_y2, b2_y2) - torch.min(b1_y1, b2_y1) # convex height 422 | if CIoU or DIoU: # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1 423 | c2 = (cw ** 2 + ch ** 2) ** alpha + eps # convex diagonal 424 | rho_x = torch.abs(b2_x1 + b2_x2 - b1_x1 - b1_x2) 425 | rho_y = torch.abs(b2_y1 + b2_y2 - b1_y1 - b1_y2) 426 | rho2 = ((rho_x ** 2 + rho_y ** 2) / 4) ** alpha # center distance 427 | if DIoU: 428 | return iou - rho2 / c2 # DIoU 429 | elif CIoU: # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47 430 | v = (4 / math.pi ** 2) * torch.pow(torch.atan(w2 / h2) - torch.atan(w1 / h1), 2) 431 | with torch.no_grad(): 432 | alpha_ciou = v / ((1 + eps) - inter / union + v) 433 | # return iou - (rho2 / c2 + v * alpha_ciou) # CIoU 434 | return iou - (rho2 / c2 + torch.pow(v * alpha_ciou + eps, alpha)) # CIoU 435 | else: # GIoU https://arxiv.org/pdf/1902.09630.pdf 436 | # c_area = cw * ch + eps # convex area 437 | # return iou - (c_area - union) / c_area # GIoU 438 | c_area = torch.max(cw * ch + eps, union) # convex area 439 | return iou - torch.pow((c_area - union) / c_area + eps, alpha) # GIoU 440 | else: 441 | return iou # torch.log(iou+eps) or iou 442 | 443 | 444 | def box_iou(box1, box2): 445 | # https://github.com/pytorch/vision/blob/master/torchvision/ops/boxes.py 446 | """ 447 | Return intersection-over-union (Jaccard index) of boxes. 448 | Both sets of boxes are expected to be in (x1, y1, x2, y2) format. 449 | Arguments: 450 | box1 (Tensor[N, 4]) 451 | box2 (Tensor[M, 4]) 452 | Returns: 453 | iou (Tensor[N, M]): the NxM matrix containing the pairwise 454 | IoU values for every element in boxes1 and boxes2 455 | """ 456 | 457 | def box_area(box): 458 | # box = 4xn 459 | return (box[2] - box[0]) * (box[3] - box[1]) 460 | 461 | area1 = box_area(box1.T) 462 | area2 = box_area(box2.T) 463 | 464 | # inter(N,M) = (rb(N,M,2) - lt(N,M,2)).clamp(0).prod(2) 465 | inter = (torch.min(box1[:, None, 2:], box2[:, 2:]) - torch.max(box1[:, None, :2], box2[:, :2])).clamp(0).prod(2) 466 | return inter / (area1[:, None] + area2 - inter) # iou = inter / (area1 + area2 - inter) 467 | 468 | 469 | def wh_iou(wh1, wh2): 470 | # Returns the nxm IoU matrix. wh1 is nx2, wh2 is mx2 471 | wh1 = wh1[:, None] # [N,1,2] 472 | wh2 = wh2[None] # [1,M,2] 473 | inter = torch.min(wh1, wh2).prod(2) # [N,M] 474 | return inter / (wh1.prod(2) + wh2.prod(2) - inter) # iou = inter / (area1 + area2 - inter) 475 | 476 | 477 | def box_giou(box1, box2): 478 | """ 479 | Return generalized intersection-over-union (Jaccard index) between two sets of boxes. 480 | Both sets of boxes are expected to be in ``(x1, y1, x2, y2)`` format with 481 | ``0 <= x1 < x2`` and ``0 <= y1 < y2``. 482 | Args: 483 | boxes1 (Tensor[N, 4]): first set of boxes 484 | boxes2 (Tensor[M, 4]): second set of boxes 485 | Returns: 486 | Tensor[N, M]: the NxM matrix containing the pairwise generalized IoU values 487 | for every element in boxes1 and boxes2 488 | """ 489 | 490 | def box_area(box): 491 | # box = 4xn 492 | return (box[2] - box[0]) * (box[3] - box[1]) 493 | 494 | area1 = box_area(box1.T) 495 | area2 = box_area(box2.T) 496 | 497 | inter = (torch.min(box1[:, None, 2:], box2[:, 2:]) - torch.max(box1[:, None, :2], box2[:, :2])).clamp(0).prod(2) 498 | union = (area1[:, None] + area2 - inter) 499 | 500 | iou = inter / union 501 | 502 | lti = torch.min(box1[:, None, :2], box2[:, :2]) 503 | rbi = torch.max(box1[:, None, 2:], box2[:, 2:]) 504 | 505 | whi = (rbi - lti).clamp(min=0) # [N,M,2] 506 | areai = whi[:, :, 0] * whi[:, :, 1] 507 | 508 | return iou - (areai - union) / areai 509 | 510 | 511 | def box_ciou(box1, box2, eps: float = 1e-7): 512 | """ 513 | Return complete intersection-over-union (Jaccard index) between two sets of boxes. 514 | Both sets of boxes are expected to be in ``(x1, y1, x2, y2)`` format with 515 | ``0 <= x1 < x2`` and ``0 <= y1 < y2``. 516 | Args: 517 | boxes1 (Tensor[N, 4]): first set of boxes 518 | boxes2 (Tensor[M, 4]): second set of boxes 519 | eps (float, optional): small number to prevent division by zero. Default: 1e-7 520 | Returns: 521 | Tensor[N, M]: the NxM matrix containing the pairwise complete IoU values 522 | for every element in boxes1 and boxes2 523 | """ 524 | 525 | def box_area(box): 526 | # box = 4xn 527 | return (box[2] - box[0]) * (box[3] - box[1]) 528 | 529 | area1 = box_area(box1.T) 530 | area2 = box_area(box2.T) 531 | 532 | inter = (torch.min(box1[:, None, 2:], box2[:, 2:]) - torch.max(box1[:, None, :2], box2[:, :2])).clamp(0).prod(2) 533 | union = (area1[:, None] + area2 - inter) 534 | 535 | iou = inter / union 536 | 537 | lti = torch.min(box1[:, None, :2], box2[:, :2]) 538 | rbi = torch.max(box1[:, None, 2:], box2[:, 2:]) 539 | 540 | whi = (rbi - lti).clamp(min=0) # [N,M,2] 541 | diagonal_distance_squared = (whi[:, :, 0] ** 2) + (whi[:, :, 1] ** 2) + eps 542 | 543 | # centers of boxes 544 | x_p = (box1[:, None, 0] + box1[:, None, 2]) / 2 545 | y_p = (box1[:, None, 1] + box1[:, None, 3]) / 2 546 | x_g = (box2[:, 0] + box2[:, 2]) / 2 547 | y_g = (box2[:, 1] + box2[:, 3]) / 2 548 | # The distance between boxes' centers squared. 549 | centers_distance_squared = (x_p - x_g) ** 2 + (y_p - y_g) ** 2 550 | 551 | w_pred = box1[:, None, 2] - box1[:, None, 0] 552 | h_pred = box1[:, None, 3] - box1[:, None, 1] 553 | 554 | w_gt = box2[:, 2] - box2[:, 0] 555 | h_gt = box2[:, 3] - box2[:, 1] 556 | 557 | v = (4 / (torch.pi ** 2)) * torch.pow((torch.atan(w_gt / h_gt) - torch.atan(w_pred / h_pred)), 2) 558 | with torch.no_grad(): 559 | alpha = v / (1 - iou + v + eps) 560 | return iou - (centers_distance_squared / diagonal_distance_squared) - alpha * v 561 | 562 | 563 | def box_diou(box1, box2, eps: float = 1e-7): 564 | """ 565 | Return distance intersection-over-union (Jaccard index) between two sets of boxes. 566 | Both sets of boxes are expected to be in ``(x1, y1, x2, y2)`` format with 567 | ``0 <= x1 < x2`` and ``0 <= y1 < y2``. 568 | Args: 569 | boxes1 (Tensor[N, 4]): first set of boxes 570 | boxes2 (Tensor[M, 4]): second set of boxes 571 | eps (float, optional): small number to prevent division by zero. Default: 1e-7 572 | Returns: 573 | Tensor[N, M]: the NxM matrix containing the pairwise distance IoU values 574 | for every element in boxes1 and boxes2 575 | """ 576 | 577 | def box_area(box): 578 | # box = 4xn 579 | return (box[2] - box[0]) * (box[3] - box[1]) 580 | 581 | area1 = box_area(box1.T) 582 | area2 = box_area(box2.T) 583 | 584 | inter = (torch.min(box1[:, None, 2:], box2[:, 2:]) - torch.max(box1[:, None, :2], box2[:, :2])).clamp(0).prod(2) 585 | union = (area1[:, None] + area2 - inter) 586 | 587 | iou = inter / union 588 | 589 | lti = torch.min(box1[:, None, :2], box2[:, :2]) 590 | rbi = torch.max(box1[:, None, 2:], box2[:, 2:]) 591 | 592 | whi = (rbi - lti).clamp(min=0) # [N,M,2] 593 | diagonal_distance_squared = (whi[:, :, 0] ** 2) + (whi[:, :, 1] ** 2) + eps 594 | 595 | # centers of boxes 596 | x_p = (box1[:, None, 0] + box1[:, None, 2]) / 2 597 | y_p = (box1[:, None, 1] + box1[:, None, 3]) / 2 598 | x_g = (box2[:, 0] + box2[:, 2]) / 2 599 | y_g = (box2[:, 1] + box2[:, 3]) / 2 600 | # The distance between boxes' centers squared. 601 | centers_distance_squared = (x_p - x_g) ** 2 + (y_p - y_g) ** 2 602 | 603 | # The distance IoU is the IoU penalized by a normalized 604 | # distance between boxes' centers squared. 605 | return iou - (centers_distance_squared / diagonal_distance_squared) 606 | 607 | 608 | def non_max_suppression(prediction, conf_thres=0.25, iou_thres=0.45, classes=None, agnostic=False, multi_label=False, 609 | labels=()): 610 | """Runs Non-Maximum Suppression (NMS) on inference results 611 | 612 | Returns: 613 | list of detections, on (n,6) tensor per image [xyxy, conf, cls] 614 | """ 615 | 616 | nc = prediction.shape[2] - 5 # number of classes 617 | xc = prediction[..., 4] > conf_thres # candidates 618 | 619 | # Settings 620 | min_wh, max_wh = 2, 4096 # (pixels) minimum and maximum box width and height 621 | max_det = 300 # maximum number of detections per image 622 | max_nms = 30000 # maximum number of boxes into torchvision.ops.nms() 623 | time_limit = 10.0 # seconds to quit after 624 | redundant = True # require redundant detections 625 | multi_label &= nc > 1 # multiple labels per box (adds 0.5ms/img) 626 | merge = False # use merge-NMS 627 | 628 | t = time.time() 629 | output = [torch.zeros((0, 6), device=prediction.device)] * prediction.shape[0] 630 | for xi, x in enumerate(prediction): # image index, image inference 631 | # Apply constraints 632 | # x[((x[..., 2:4] < min_wh) | (x[..., 2:4] > max_wh)).any(1), 4] = 0 # width-height 633 | x = x[xc[xi]] # confidence 634 | 635 | # Cat apriori labels if autolabelling 636 | if labels and len(labels[xi]): 637 | l = labels[xi] 638 | v = torch.zeros((len(l), nc + 5), device=x.device) 639 | v[:, :4] = l[:, 1:5] # box 640 | v[:, 4] = 1.0 # conf 641 | v[range(len(l)), l[:, 0].long() + 5] = 1.0 # cls 642 | x = torch.cat((x, v), 0) 643 | 644 | # If none remain process next image 645 | if not x.shape[0]: 646 | continue 647 | 648 | # Compute conf 649 | if nc == 1: 650 | x[:, 5:] = x[:, 4:5] # for models with one class, cls_loss is 0 and cls_conf is always 0.5, 651 | # so there is no need to multiplicate. 652 | else: 653 | x[:, 5:] *= x[:, 4:5] # conf = obj_conf * cls_conf 654 | 655 | # Box (center x, center y, width, height) to (x1, y1, x2, y2) 656 | box = xywh2xyxy(x[:, :4]) 657 | 658 | # Detections matrix nx6 (xyxy, conf, cls) 659 | if multi_label: 660 | i, j = (x[:, 5:] > conf_thres).nonzero(as_tuple=False).T 661 | x = torch.cat((box[i], x[i, j + 5, None], j[:, None].float()), 1) 662 | else: # best class only 663 | conf, j = x[:, 5:].max(1, keepdim=True) 664 | x = torch.cat((box, conf, j.float()), 1)[conf.view(-1) > conf_thres] 665 | 666 | # Filter by class 667 | if classes is not None: 668 | x = x[(x[:, 5:6] == torch.tensor(classes, device=x.device)).any(1)] 669 | 670 | # Apply finite constraint 671 | # if not torch.isfinite(x).all(): 672 | # x = x[torch.isfinite(x).all(1)] 673 | 674 | # Check shape 675 | n = x.shape[0] # number of boxes 676 | if not n: # no boxes 677 | continue 678 | elif n > max_nms: # excess boxes 679 | x = x[x[:, 4].argsort(descending=True)[:max_nms]] # sort by confidence 680 | 681 | # Batched NMS 682 | c = x[:, 5:6] * (0 if agnostic else max_wh) # classes 683 | boxes, scores = x[:, :4] + c, x[:, 4] # boxes (offset by class), scores 684 | i = torchvision.ops.nms(boxes, scores, iou_thres) # NMS 685 | if i.shape[0] > max_det: # limit detections 686 | i = i[:max_det] 687 | if merge and (1 < n < 3E3): # Merge NMS (boxes merged using weighted mean) 688 | # update boxes as boxes(i,4) = weights(i,n) * boxes(n,4) 689 | iou = box_iou(boxes[i], boxes) > iou_thres # iou matrix 690 | weights = iou * scores[None] # box weights 691 | x[i, :4] = torch.mm(weights, x[:, :4]).float() / weights.sum(1, keepdim=True) # merged boxes 692 | if redundant: 693 | i = i[iou.sum(1) > 1] # require redundancy 694 | 695 | output[xi] = x[i] 696 | if (time.time() - t) > time_limit: 697 | print(f'WARNING: NMS time limit {time_limit}s exceeded') 698 | break # time limit exceeded 699 | 700 | return output 701 | 702 | 703 | def non_max_suppression_kpt(prediction, conf_thres=0.25, iou_thres=0.45, classes=None, agnostic=False, multi_label=False, 704 | labels=(), kpt_label=False, nc=None, nkpt=None): 705 | """Runs Non-Maximum Suppression (NMS) on inference results 706 | 707 | Returns: 708 | list of detections, on (n,6) tensor per image [xyxy, conf, cls] 709 | """ 710 | if nc is None: 711 | nc = prediction.shape[2] - 5 if not kpt_label else prediction.shape[2] - 56 # number of classes 712 | xc = prediction[..., 4] > conf_thres # candidates 713 | 714 | # Settings 715 | min_wh, max_wh = 2, 4096 # (pixels) minimum and maximum box width and height 716 | max_det = 300 # maximum number of detections per image 717 | max_nms = 30000 # maximum number of boxes into torchvision.ops.nms() 718 | time_limit = 10.0 # seconds to quit after 719 | redundant = True # require redundant detections 720 | multi_label &= nc > 1 # multiple labels per box (adds 0.5ms/img) 721 | merge = False # use merge-NMS 722 | 723 | t = time.time() 724 | output = [torch.zeros((0,6), device=prediction.device)] * prediction.shape[0] 725 | for xi, x in enumerate(prediction): # image index, image inference 726 | # Apply constraints 727 | # x[((x[..., 2:4] < min_wh) | (x[..., 2:4] > max_wh)).any(1), 4] = 0 # width-height 728 | x = x[xc[xi]] # confidence 729 | 730 | # Cat apriori labels if autolabelling 731 | if labels and len(labels[xi]): 732 | l = labels[xi] 733 | v = torch.zeros((len(l), nc + 5), device=x.device) 734 | v[:, :4] = l[:, 1:5] # box 735 | v[:, 4] = 1.0 # conf 736 | v[range(len(l)), l[:, 0].long() + 5] = 1.0 # cls 737 | x = torch.cat((x, v), 0) 738 | 739 | # If none remain process next image 740 | if not x.shape[0]: 741 | continue 742 | 743 | # Compute conf 744 | x[:, 5:5+nc] *= x[:, 4:5] # conf = obj_conf * cls_conf 745 | 746 | # Box (center x, center y, width, height) to (x1, y1, x2, y2) 747 | box = xywh2xyxy(x[:, :4]) 748 | 749 | # Detections matrix nx6 (xyxy, conf, cls) 750 | if multi_label: 751 | i, j = (x[:, 5:] > conf_thres).nonzero(as_tuple=False).T 752 | x = torch.cat((box[i], x[i, j + 5, None], j[:, None].float()), 1) 753 | else: # best class only 754 | if not kpt_label: 755 | conf, j = x[:, 5:].max(1, keepdim=True) 756 | x = torch.cat((box, conf, j.float()), 1)[conf.view(-1) > conf_thres] 757 | else: 758 | kpts = x[:, 6:] 759 | conf, j = x[:, 5:6].max(1, keepdim=True) 760 | x = torch.cat((box, conf, j.float(), kpts), 1)[conf.view(-1) > conf_thres] 761 | 762 | 763 | # Filter by class 764 | if classes is not None: 765 | x = x[(x[:, 5:6] == torch.tensor(classes, device=x.device)).any(1)] 766 | 767 | # Apply finite constraint 768 | # if not torch.isfinite(x).all(): 769 | # x = x[torch.isfinite(x).all(1)] 770 | 771 | # Check shape 772 | n = x.shape[0] # number of boxes 773 | if not n: # no boxes 774 | continue 775 | elif n > max_nms: # excess boxes 776 | x = x[x[:, 4].argsort(descending=True)[:max_nms]] # sort by confidence 777 | 778 | # Batched NMS 779 | c = x[:, 5:6] * (0 if agnostic else max_wh) # classes 780 | boxes, scores = x[:, :4] + c, x[:, 4] # boxes (offset by class), scores 781 | i = torchvision.ops.nms(boxes, scores, iou_thres) # NMS 782 | if i.shape[0] > max_det: # limit detections 783 | i = i[:max_det] 784 | if merge and (1 < n < 3E3): # Merge NMS (boxes merged using weighted mean) 785 | # update boxes as boxes(i,4) = weights(i,n) * boxes(n,4) 786 | iou = box_iou(boxes[i], boxes) > iou_thres # iou matrix 787 | weights = iou * scores[None] # box weights 788 | x[i, :4] = torch.mm(weights, x[:, :4]).float() / weights.sum(1, keepdim=True) # merged boxes 789 | if redundant: 790 | i = i[iou.sum(1) > 1] # require redundancy 791 | 792 | output[xi] = x[i] 793 | if (time.time() - t) > time_limit: 794 | print(f'WARNING: NMS time limit {time_limit}s exceeded') 795 | break # time limit exceeded 796 | 797 | return output 798 | 799 | 800 | def strip_optimizer(f='best.pt', s=''): # from utils.general import *; strip_optimizer() 801 | # Strip optimizer from 'f' to finalize training, optionally save as 's' 802 | x = torch.load(f, map_location=torch.device('cpu')) 803 | if x.get('ema'): 804 | x['model'] = x['ema'] # replace model with ema 805 | for k in 'optimizer', 'training_results', 'wandb_id', 'ema', 'updates': # keys 806 | x[k] = None 807 | x['epoch'] = -1 808 | x['model'].half() # to FP16 809 | for p in x['model'].parameters(): 810 | p.requires_grad = False 811 | torch.save(x, s or f) 812 | mb = os.path.getsize(s or f) / 1E6 # filesize 813 | print(f"Optimizer stripped from {f},{(' saved as %s,' % s) if s else ''} {mb:.1f}MB") 814 | 815 | 816 | def print_mutation(hyp, results, yaml_file='hyp_evolved.yaml', bucket=''): 817 | # Print mutation results to evolve.txt (for use with train.py --evolve) 818 | a = '%10s' * len(hyp) % tuple(hyp.keys()) # hyperparam keys 819 | b = '%10.3g' * len(hyp) % tuple(hyp.values()) # hyperparam values 820 | c = '%10.4g' * len(results) % results # results (P, R, mAP@0.5, mAP@0.5:0.95, val_losses x 3) 821 | print('\n%s\n%s\nEvolved fitness: %s\n' % (a, b, c)) 822 | 823 | if bucket: 824 | url = 'gs://%s/evolve.txt' % bucket 825 | if gsutil_getsize(url) > (os.path.getsize('evolve.txt') if os.path.exists('evolve.txt') else 0): 826 | os.system('gsutil cp %s .' % url) # download evolve.txt if larger than local 827 | 828 | with open('evolve.txt', 'a') as f: # append result 829 | f.write(c + b + '\n') 830 | x = np.unique(np.loadtxt('evolve.txt', ndmin=2), axis=0) # load unique rows 831 | x = x[np.argsort(-fitness(x))] # sort 832 | np.savetxt('evolve.txt', x, '%10.3g') # save sort by fitness 833 | 834 | # Save yaml 835 | for i, k in enumerate(hyp.keys()): 836 | hyp[k] = float(x[0, i + 7]) 837 | with open(yaml_file, 'w') as f: 838 | results = tuple(x[0, :7]) 839 | c = '%10.4g' * len(results) % results # results (P, R, mAP@0.5, mAP@0.5:0.95, val_losses x 3) 840 | f.write('# Hyperparameter Evolution Results\n# Generations: %g\n# Metrics: ' % len(x) + c + '\n\n') 841 | yaml.dump(hyp, f, sort_keys=False) 842 | 843 | if bucket: 844 | os.system('gsutil cp evolve.txt %s gs://%s' % (yaml_file, bucket)) # upload 845 | 846 | 847 | def apply_classifier(x, model, img, im0): 848 | # applies a second stage classifier to yolo outputs 849 | im0 = [im0] if isinstance(im0, np.ndarray) else im0 850 | for i, d in enumerate(x): # per image 851 | if d is not None and len(d): 852 | d = d.clone() 853 | 854 | # Reshape and pad cutouts 855 | b = xyxy2xywh(d[:, :4]) # boxes 856 | b[:, 2:] = b[:, 2:].max(1)[0].unsqueeze(1) # rectangle to square 857 | b[:, 2:] = b[:, 2:] * 1.3 + 30 # pad 858 | d[:, :4] = xywh2xyxy(b).long() 859 | 860 | # Rescale boxes from img_size to im0 size 861 | scale_coords(img.shape[2:], d[:, :4], im0[i].shape) 862 | 863 | # Classes 864 | pred_cls1 = d[:, 5].long() 865 | ims = [] 866 | for j, a in enumerate(d): # per item 867 | cutout = im0[i][int(a[1]):int(a[3]), int(a[0]):int(a[2])] 868 | im = cv2.resize(cutout, (224, 224)) # BGR 869 | # cv2.imwrite('test%i.jpg' % j, cutout) 870 | 871 | im = im[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416 872 | im = np.ascontiguousarray(im, dtype=np.float32) # uint8 to float32 873 | im /= 255.0 # 0 - 255 to 0.0 - 1.0 874 | ims.append(im) 875 | 876 | pred_cls2 = model(torch.Tensor(ims).to(d.device)).argmax(1) # classifier prediction 877 | x[i] = x[i][pred_cls1 == pred_cls2] # retain matching class detections 878 | 879 | return x 880 | 881 | 882 | def increment_path(path, exist_ok=True, sep=''): 883 | # Increment path, i.e. runs/exp --> runs/exp{sep}0, runs/exp{sep}1 etc. 884 | path = Path(path) # os-agnostic 885 | if (path.exists() and exist_ok) or (not path.exists()): 886 | return str(path) 887 | else: 888 | dirs = glob.glob(f"{path}{sep}*") # similar paths 889 | matches = [re.search(rf"%s{sep}(\d+)" % path.stem, d) for d in dirs] 890 | i = [int(m.groups()[0]) for m in matches if m] # indices 891 | n = max(i) + 1 if i else 2 # increment number 892 | return f"{path}{sep}{n}" # update path 893 | -------------------------------------------------------------------------------- /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 | -------------------------------------------------------------------------------- /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 | -------------------------------------------------------------------------------- /utils/google_app_engine/app.yaml: -------------------------------------------------------------------------------- 1 | runtime: custom 2 | env: flex 3 | 4 | service: yolorapp 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 -------------------------------------------------------------------------------- /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='WongKinYiu/yolov7'): 20 | # Attempt file download if does not exist 21 | file = Path(str(file).strip().replace("'", '').lower()) 22 | 23 | if not file.exists(): 24 | try: 25 | response = requests.get(f'https://api.github.com/repos/{repo}/releases/latest').json() # github api 26 | assets = [x['name'] for x in response['assets']] # release assets 27 | tag = response['tag_name'] # i.e. 'v1.0' 28 | except: # fallback plan 29 | assets = ['yolov7.pt', 'yolov7-tiny.pt', 'yolov7x.pt', 'yolov7-d6.pt', 'yolov7-e6.pt', 30 | 'yolov7-e6e.pt', 'yolov7-w6.pt'] 31 | tag = subprocess.check_output('git tag', shell=True).decode().split()[-1] 32 | 33 | name = file.name 34 | if name in assets: 35 | msg = f'{file} missing, try downloading from https://github.com/{repo}/releases/' 36 | redundant = False # second download option 37 | try: # GitHub 38 | url = f'https://github.com/{repo}/releases/download/{tag}/{name}' 39 | print(f'Downloading {url} to {file}...') 40 | torch.hub.download_url_to_file(url, file) 41 | assert file.exists() and file.stat().st_size > 1E6 # check 42 | except Exception as e: # GCP 43 | print(f'Download error: {e}') 44 | assert redundant, 'No secondary mirror' 45 | url = f'https://storage.googleapis.com/{repo}/ckpt/{name}' 46 | print(f'Downloading {url} to {file}...') 47 | os.system(f'curl -L {url} -o {file}') # torch.hub.download_url_to_file(url, weights) 48 | finally: 49 | if not file.exists() or file.stat().st_size < 1E6: # check 50 | file.unlink(missing_ok=True) # remove partial downloads 51 | print(f'ERROR: Download failure: {msg}') 52 | print('') 53 | return 54 | 55 | 56 | def gdrive_download(id='', file='tmp.zip'): 57 | # Downloads a file from Google Drive. from yolov7.utils.google_utils import *; gdrive_download() 58 | t = time.time() 59 | file = Path(file) 60 | cookie = Path('cookie') # gdrive cookie 61 | print(f'Downloading https://drive.google.com/uc?export=download&id={id} as {file}... ', end='') 62 | file.unlink(missing_ok=True) # remove existing file 63 | cookie.unlink(missing_ok=True) # remove existing cookie 64 | 65 | # Attempt file download 66 | out = "NUL" if platform.system() == "Windows" else "/dev/null" 67 | os.system(f'curl -c ./cookie -s -L "drive.google.com/uc?export=download&id={id}" > {out}') 68 | if os.path.exists('cookie'): # large file 69 | s = f'curl -Lb ./cookie "drive.google.com/uc?export=download&confirm={get_token()}&id={id}" -o {file}' 70 | else: # small file 71 | s = f'curl -s -L -o {file} "drive.google.com/uc?export=download&id={id}"' 72 | r = os.system(s) # execute, capture return 73 | cookie.unlink(missing_ok=True) # remove existing cookie 74 | 75 | # Error check 76 | if r != 0: 77 | file.unlink(missing_ok=True) # remove partial 78 | print('Download error ') # raise Exception('Download error') 79 | return r 80 | 81 | # Unzip if archive 82 | if file.suffix == '.zip': 83 | print('unzipping... ', end='') 84 | os.system(f'unzip -q {file}') # unzip 85 | file.unlink() # remove zip to free space 86 | 87 | print(f'Done ({time.time() - t:.1f}s)') 88 | return r 89 | 90 | 91 | def get_token(cookie="./cookie"): 92 | with open(cookie) as f: 93 | for line in f: 94 | if "download" in line: 95 | return line.split()[-1] 96 | return "" 97 | 98 | # def upload_blob(bucket_name, source_file_name, destination_blob_name): 99 | # # Uploads a file to a bucket 100 | # # https://cloud.google.com/storage/docs/uploading-objects#storage-upload-object-python 101 | # 102 | # storage_client = storage.Client() 103 | # bucket = storage_client.get_bucket(bucket_name) 104 | # blob = bucket.blob(destination_blob_name) 105 | # 106 | # blob.upload_from_filename(source_file_name) 107 | # 108 | # print('File {} uploaded to {}.'.format( 109 | # source_file_name, 110 | # destination_blob_name)) 111 | # 112 | # 113 | # def download_blob(bucket_name, source_blob_name, destination_file_name): 114 | # # Uploads a blob from a bucket 115 | # storage_client = storage.Client() 116 | # bucket = storage_client.get_bucket(bucket_name) 117 | # blob = bucket.blob(source_blob_name) 118 | # 119 | # blob.download_to_filename(destination_file_name) 120 | # 121 | # print('Blob {} downloaded to {}.'.format( 122 | # source_blob_name, 123 | # destination_file_name)) 124 | -------------------------------------------------------------------------------- /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, v5_metric=False, 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], v5_metric=v5_metric) 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, v5_metric=False): 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 | v5_metric: Assume maximum recall to be 1.0, as in YOLOv5, MMDetetion etc. 87 | # Returns 88 | Average precision, precision curve, recall curve 89 | """ 90 | 91 | # Append sentinel values to beginning and end 92 | if v5_metric: # New YOLOv5 metric, same as MMDetection and Detectron2 repositories 93 | mrec = np.concatenate(([0.], recall, [1.0])) 94 | else: # Old YOLOv5 metric, i.e. default YOLOv7 metric 95 | mrec = np.concatenate(([0.], recall, [recall[-1] + 0.01])) 96 | mpre = np.concatenate(([1.], precision, [0.])) 97 | 98 | # Compute the precision envelope 99 | mpre = np.flip(np.maximum.accumulate(np.flip(mpre))) 100 | 101 | # Integrate area under curve 102 | method = 'interp' # methods: 'continuous', 'interp' 103 | if method == 'interp': 104 | x = np.linspace(0, 1, 101) # 101-point interp (COCO) 105 | ap = np.trapz(np.interp(x, mrec, mpre), x) # integrate 106 | else: # 'continuous' 107 | i = np.where(mrec[1:] != mrec[:-1])[0] # points where x axis (recall) changes 108 | ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1]) # area under curve 109 | 110 | return ap, mpre, mrec 111 | 112 | 113 | class ConfusionMatrix: 114 | # Updated version of https://github.com/kaanakan/object_detection_confusion_matrix 115 | def __init__(self, nc, conf=0.25, iou_thres=0.45): 116 | self.matrix = np.zeros((nc + 1, nc + 1)) 117 | self.nc = nc # number of classes 118 | self.conf = conf 119 | self.iou_thres = iou_thres 120 | 121 | def process_batch(self, detections, labels): 122 | """ 123 | Return intersection-over-union (Jaccard index) of boxes. 124 | Both sets of boxes are expected to be in (x1, y1, x2, y2) format. 125 | Arguments: 126 | detections (Array[N, 6]), x1, y1, x2, y2, conf, class 127 | labels (Array[M, 5]), class, x1, y1, x2, y2 128 | Returns: 129 | None, updates confusion matrix accordingly 130 | """ 131 | detections = detections[detections[:, 4] > self.conf] 132 | gt_classes = labels[:, 0].int() 133 | detection_classes = detections[:, 5].int() 134 | iou = general.box_iou(labels[:, 1:], detections[:, :4]) 135 | 136 | x = torch.where(iou > self.iou_thres) 137 | if x[0].shape[0]: 138 | matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy() 139 | if x[0].shape[0] > 1: 140 | matches = matches[matches[:, 2].argsort()[::-1]] 141 | matches = matches[np.unique(matches[:, 1], return_index=True)[1]] 142 | matches = matches[matches[:, 2].argsort()[::-1]] 143 | matches = matches[np.unique(matches[:, 0], return_index=True)[1]] 144 | else: 145 | matches = np.zeros((0, 3)) 146 | 147 | n = matches.shape[0] > 0 148 | m0, m1, _ = matches.transpose().astype(np.int16) 149 | for i, gc in enumerate(gt_classes): 150 | j = m0 == i 151 | if n and sum(j) == 1: 152 | self.matrix[gc, detection_classes[m1[j]]] += 1 # correct 153 | else: 154 | self.matrix[self.nc, gc] += 1 # background FP 155 | 156 | if n: 157 | for i, dc in enumerate(detection_classes): 158 | if not any(m1 == i): 159 | self.matrix[dc, self.nc] += 1 # background FN 160 | 161 | def matrix(self): 162 | return self.matrix 163 | 164 | def plot(self, save_dir='', names=()): 165 | try: 166 | import seaborn as sn 167 | 168 | array = self.matrix / (self.matrix.sum(0).reshape(1, self.nc + 1) + 1E-6) # normalize 169 | array[array < 0.005] = np.nan # don't annotate (would appear as 0.00) 170 | 171 | fig = plt.figure(figsize=(12, 9), tight_layout=True) 172 | sn.set(font_scale=1.0 if self.nc < 50 else 0.8) # for label size 173 | labels = (0 < len(names) < 99) and len(names) == self.nc # apply names to ticklabels 174 | sn.heatmap(array, annot=self.nc < 30, annot_kws={"size": 8}, cmap='Blues', fmt='.2f', square=True, 175 | xticklabels=names + ['background FP'] if labels else "auto", 176 | yticklabels=names + ['background FN'] if labels else "auto").set_facecolor((1, 1, 1)) 177 | fig.axes[0].set_xlabel('True') 178 | fig.axes[0].set_ylabel('Predicted') 179 | fig.savefig(Path(save_dir) / 'confusion_matrix.png', dpi=250) 180 | except Exception as e: 181 | pass 182 | 183 | def print(self): 184 | for i in range(self.nc + 1): 185 | print(' '.join(map(str, self.matrix[i]))) 186 | 187 | 188 | # Plots ---------------------------------------------------------------------------------------------------------------- 189 | 190 | def plot_pr_curve(px, py, ap, save_dir='pr_curve.png', names=()): 191 | # Precision-recall curve 192 | fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True) 193 | py = np.stack(py, axis=1) 194 | 195 | if 0 < len(names) < 21: # display per-class legend if < 21 classes 196 | for i, y in enumerate(py.T): 197 | ax.plot(px, y, linewidth=1, label=f'{names[i]} {ap[i, 0]:.3f}') # plot(recall, precision) 198 | else: 199 | ax.plot(px, py, linewidth=1, color='grey') # plot(recall, precision) 200 | 201 | ax.plot(px, py.mean(1), linewidth=3, color='blue', label='all classes %.3f mAP@0.5' % ap[:, 0].mean()) 202 | ax.set_xlabel('Recall') 203 | ax.set_ylabel('Precision') 204 | ax.set_xlim(0, 1) 205 | ax.set_ylim(0, 1) 206 | plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left") 207 | fig.savefig(Path(save_dir), dpi=250) 208 | 209 | 210 | def plot_mc_curve(px, py, save_dir='mc_curve.png', names=(), xlabel='Confidence', ylabel='Metric'): 211 | # Metric-confidence curve 212 | fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True) 213 | 214 | if 0 < len(names) < 21: # display per-class legend if < 21 classes 215 | for i, y in enumerate(py): 216 | ax.plot(px, y, linewidth=1, label=f'{names[i]}') # plot(confidence, metric) 217 | else: 218 | ax.plot(px, py.T, linewidth=1, color='grey') # plot(confidence, metric) 219 | 220 | y = py.mean(0) 221 | ax.plot(px, y, linewidth=3, color='blue', label=f'all classes {y.max():.2f} at {px[y.argmax()]:.3f}') 222 | ax.set_xlabel(xlabel) 223 | ax.set_ylabel(ylabel) 224 | ax.set_xlim(0, 1) 225 | ax.set_ylim(0, 1) 226 | plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left") 227 | fig.savefig(Path(save_dir), dpi=250) 228 | -------------------------------------------------------------------------------- /utils/plots.py: -------------------------------------------------------------------------------- 1 | # Plotting utils 2 | 3 | import glob 4 | import math 5 | import os 6 | import random 7 | from copy import copy 8 | from pathlib import Path 9 | 10 | import cv2 11 | import matplotlib 12 | import matplotlib.pyplot as plt 13 | import numpy as np 14 | import pandas as pd 15 | import seaborn as sns 16 | import torch 17 | import yaml 18 | from PIL import Image, ImageDraw, ImageFont 19 | from scipy.signal import butter, filtfilt 20 | 21 | from utils.general import xywh2xyxy, xyxy2xywh 22 | from utils.metrics import fitness 23 | 24 | # Settings 25 | matplotlib.rc('font', **{'size': 11}) 26 | matplotlib.use('Agg') # for writing to files only 27 | 28 | 29 | def color_list(): 30 | # Return first 10 plt colors as (r,g,b) https://stackoverflow.com/questions/51350872/python-from-color-name-to-rgb 31 | def hex2rgb(h): 32 | return tuple(int(h[1 + i:1 + i + 2], 16) for i in (0, 2, 4)) 33 | 34 | return [hex2rgb(h) for h in matplotlib.colors.TABLEAU_COLORS.values()] # or BASE_ (8), CSS4_ (148), XKCD_ (949) 35 | 36 | 37 | def hist2d(x, y, n=100): 38 | # 2d histogram used in labels.png and evolve.png 39 | xedges, yedges = np.linspace(x.min(), x.max(), n), np.linspace(y.min(), y.max(), n) 40 | hist, xedges, yedges = np.histogram2d(x, y, (xedges, yedges)) 41 | xidx = np.clip(np.digitize(x, xedges) - 1, 0, hist.shape[0] - 1) 42 | yidx = np.clip(np.digitize(y, yedges) - 1, 0, hist.shape[1] - 1) 43 | return np.log(hist[xidx, yidx]) 44 | 45 | 46 | def butter_lowpass_filtfilt(data, cutoff=1500, fs=50000, order=5): 47 | # https://stackoverflow.com/questions/28536191/how-to-filter-smooth-with-scipy-numpy 48 | def butter_lowpass(cutoff, fs, order): 49 | nyq = 0.5 * fs 50 | normal_cutoff = cutoff / nyq 51 | return butter(order, normal_cutoff, btype='low', analog=False) 52 | 53 | b, a = butter_lowpass(cutoff, fs, order=order) 54 | return filtfilt(b, a, data) # forward-backward filter 55 | 56 | 57 | def plot_one_box(x, img, color=None, label=None, line_thickness=3): 58 | # Plots one bounding box on image img 59 | tl = line_thickness or round(0.002 * (img.shape[0] + img.shape[1]) / 2) + 1 # line/font thickness 60 | color = color or [random.randint(0, 255) for _ in range(3)] 61 | c1, c2 = (int(x[0]), int(x[1])), (int(x[2]), int(x[3])) 62 | cv2.rectangle(img, c1, c2, color, thickness=tl, lineType=cv2.LINE_AA) 63 | if label: 64 | tf = max(tl - 1, 1) # font thickness 65 | t_size = cv2.getTextSize(label, 0, fontScale=tl / 3, thickness=tf)[0] 66 | c2 = c1[0] + t_size[0], c1[1] - t_size[1] - 3 67 | cv2.rectangle(img, c1, c2, color, -1, cv2.LINE_AA) # filled 68 | cv2.putText(img, label, (c1[0], c1[1] - 2), 0, tl / 3, [225, 255, 255], thickness=tf, lineType=cv2.LINE_AA) 69 | 70 | 71 | def plot_one_box_PIL(box, img, color=None, label=None, line_thickness=None): 72 | img = Image.fromarray(img) 73 | draw = ImageDraw.Draw(img) 74 | line_thickness = line_thickness or max(int(min(img.size) / 200), 2) 75 | draw.rectangle(box, width=line_thickness, outline=tuple(color)) # plot 76 | if label: 77 | fontsize = max(round(max(img.size) / 40), 12) 78 | font = ImageFont.truetype("Arial.ttf", fontsize) 79 | txt_width, txt_height = font.getsize(label) 80 | draw.rectangle([box[0], box[1] - txt_height + 4, box[0] + txt_width, box[1]], fill=tuple(color)) 81 | draw.text((box[0], box[1] - txt_height + 1), label, fill=(255, 255, 255), font=font) 82 | return np.asarray(img) 83 | 84 | 85 | def plot_wh_methods(): # from utils.plots import *; plot_wh_methods() 86 | # Compares the two methods for width-height anchor multiplication 87 | # https://github.com/ultralytics/yolov3/issues/168 88 | x = np.arange(-4.0, 4.0, .1) 89 | ya = np.exp(x) 90 | yb = torch.sigmoid(torch.from_numpy(x)).numpy() * 2 91 | 92 | fig = plt.figure(figsize=(6, 3), tight_layout=True) 93 | plt.plot(x, ya, '.-', label='YOLOv3') 94 | plt.plot(x, yb ** 2, '.-', label='YOLOR ^2') 95 | plt.plot(x, yb ** 1.6, '.-', label='YOLOR ^1.6') 96 | plt.xlim(left=-4, right=4) 97 | plt.ylim(bottom=0, top=6) 98 | plt.xlabel('input') 99 | plt.ylabel('output') 100 | plt.grid() 101 | plt.legend() 102 | fig.savefig('comparison.png', dpi=200) 103 | 104 | 105 | def output_to_target(output): 106 | # Convert model output to target format [batch_id, class_id, x, y, w, h, conf] 107 | targets = [] 108 | for i, o in enumerate(output): 109 | for *box, conf, cls in o.cpu().numpy(): 110 | targets.append([i, cls, *list(*xyxy2xywh(np.array(box)[None])), conf]) 111 | return np.array(targets) 112 | 113 | 114 | def plot_images(images, targets, paths=None, fname='images.jpg', names=None, max_size=640, max_subplots=16): 115 | # Plot image grid with labels 116 | 117 | if isinstance(images, torch.Tensor): 118 | images = images.cpu().float().numpy() 119 | if isinstance(targets, torch.Tensor): 120 | targets = targets.cpu().numpy() 121 | 122 | # un-normalise 123 | if np.max(images[0]) <= 1: 124 | images *= 255 125 | 126 | tl = 3 # line thickness 127 | tf = max(tl - 1, 1) # font thickness 128 | bs, _, h, w = images.shape # batch size, _, height, width 129 | bs = min(bs, max_subplots) # limit plot images 130 | ns = np.ceil(bs ** 0.5) # number of subplots (square) 131 | 132 | # Check if we should resize 133 | scale_factor = max_size / max(h, w) 134 | if scale_factor < 1: 135 | h = math.ceil(scale_factor * h) 136 | w = math.ceil(scale_factor * w) 137 | 138 | colors = color_list() # list of colors 139 | mosaic = np.full((int(ns * h), int(ns * w), 3), 255, dtype=np.uint8) # init 140 | for i, img in enumerate(images): 141 | if i == max_subplots: # if last batch has fewer images than we expect 142 | break 143 | 144 | block_x = int(w * (i // ns)) 145 | block_y = int(h * (i % ns)) 146 | 147 | img = img.transpose(1, 2, 0) 148 | if scale_factor < 1: 149 | img = cv2.resize(img, (w, h)) 150 | 151 | mosaic[block_y:block_y + h, block_x:block_x + w, :] = img 152 | if len(targets) > 0: 153 | image_targets = targets[targets[:, 0] == i] 154 | boxes = xywh2xyxy(image_targets[:, 2:6]).T 155 | classes = image_targets[:, 1].astype('int') 156 | labels = image_targets.shape[1] == 6 # labels if no conf column 157 | conf = None if labels else image_targets[:, 6] # check for confidence presence (label vs pred) 158 | 159 | if boxes.shape[1]: 160 | if boxes.max() <= 1.01: # if normalized with tolerance 0.01 161 | boxes[[0, 2]] *= w # scale to pixels 162 | boxes[[1, 3]] *= h 163 | elif scale_factor < 1: # absolute coords need scale if image scales 164 | boxes *= scale_factor 165 | boxes[[0, 2]] += block_x 166 | boxes[[1, 3]] += block_y 167 | for j, box in enumerate(boxes.T): 168 | cls = int(classes[j]) 169 | color = colors[cls % len(colors)] 170 | cls = names[cls] if names else cls 171 | if labels or conf[j] > 0.25: # 0.25 conf thresh 172 | label = '%s' % cls if labels else '%s %.1f' % (cls, conf[j]) 173 | plot_one_box(box, mosaic, label=label, color=color, line_thickness=tl) 174 | 175 | # Draw image filename labels 176 | if paths: 177 | label = Path(paths[i]).name[:40] # trim to 40 char 178 | t_size = cv2.getTextSize(label, 0, fontScale=tl / 3, thickness=tf)[0] 179 | cv2.putText(mosaic, label, (block_x + 5, block_y + t_size[1] + 5), 0, tl / 3, [220, 220, 220], thickness=tf, 180 | lineType=cv2.LINE_AA) 181 | 182 | # Image border 183 | cv2.rectangle(mosaic, (block_x, block_y), (block_x + w, block_y + h), (255, 255, 255), thickness=3) 184 | 185 | if fname: 186 | r = min(1280. / max(h, w) / ns, 1.0) # ratio to limit image size 187 | mosaic = cv2.resize(mosaic, (int(ns * w * r), int(ns * h * r)), interpolation=cv2.INTER_AREA) 188 | # cv2.imwrite(fname, cv2.cvtColor(mosaic, cv2.COLOR_BGR2RGB)) # cv2 save 189 | Image.fromarray(mosaic).save(fname) # PIL save 190 | return mosaic 191 | 192 | 193 | def plot_lr_scheduler(optimizer, scheduler, epochs=300, save_dir=''): 194 | # Plot LR simulating training for full epochs 195 | optimizer, scheduler = copy(optimizer), copy(scheduler) # do not modify originals 196 | y = [] 197 | for _ in range(epochs): 198 | scheduler.step() 199 | y.append(optimizer.param_groups[0]['lr']) 200 | plt.plot(y, '.-', label='LR') 201 | plt.xlabel('epoch') 202 | plt.ylabel('LR') 203 | plt.grid() 204 | plt.xlim(0, epochs) 205 | plt.ylim(0) 206 | plt.savefig(Path(save_dir) / 'LR.png', dpi=200) 207 | plt.close() 208 | 209 | 210 | def plot_test_txt(): # from utils.plots import *; plot_test() 211 | # Plot test.txt histograms 212 | x = np.loadtxt('test.txt', dtype=np.float32) 213 | box = xyxy2xywh(x[:, :4]) 214 | cx, cy = box[:, 0], box[:, 1] 215 | 216 | fig, ax = plt.subplots(1, 1, figsize=(6, 6), tight_layout=True) 217 | ax.hist2d(cx, cy, bins=600, cmax=10, cmin=0) 218 | ax.set_aspect('equal') 219 | plt.savefig('hist2d.png', dpi=300) 220 | 221 | fig, ax = plt.subplots(1, 2, figsize=(12, 6), tight_layout=True) 222 | ax[0].hist(cx, bins=600) 223 | ax[1].hist(cy, bins=600) 224 | plt.savefig('hist1d.png', dpi=200) 225 | 226 | 227 | def plot_targets_txt(): # from utils.plots import *; plot_targets_txt() 228 | # Plot targets.txt histograms 229 | x = np.loadtxt('targets.txt', dtype=np.float32).T 230 | s = ['x targets', 'y targets', 'width targets', 'height targets'] 231 | fig, ax = plt.subplots(2, 2, figsize=(8, 8), tight_layout=True) 232 | ax = ax.ravel() 233 | for i in range(4): 234 | ax[i].hist(x[i], bins=100, label='%.3g +/- %.3g' % (x[i].mean(), x[i].std())) 235 | ax[i].legend() 236 | ax[i].set_title(s[i]) 237 | plt.savefig('targets.jpg', dpi=200) 238 | 239 | 240 | def plot_study_txt(path='', x=None): # from utils.plots import *; plot_study_txt() 241 | # Plot study.txt generated by test.py 242 | fig, ax = plt.subplots(2, 4, figsize=(10, 6), tight_layout=True) 243 | # ax = ax.ravel() 244 | 245 | fig2, ax2 = plt.subplots(1, 1, figsize=(8, 4), tight_layout=True) 246 | # for f in [Path(path) / f'study_coco_{x}.txt' for x in ['yolor-p6', 'yolor-w6', 'yolor-e6', 'yolor-d6']]: 247 | for f in sorted(Path(path).glob('study*.txt')): 248 | y = np.loadtxt(f, dtype=np.float32, usecols=[0, 1, 2, 3, 7, 8, 9], ndmin=2).T 249 | x = np.arange(y.shape[1]) if x is None else np.array(x) 250 | s = ['P', 'R', 'mAP@.5', 'mAP@.5:.95', 't_inference (ms/img)', 't_NMS (ms/img)', 't_total (ms/img)'] 251 | # for i in range(7): 252 | # ax[i].plot(x, y[i], '.-', linewidth=2, markersize=8) 253 | # ax[i].set_title(s[i]) 254 | 255 | j = y[3].argmax() + 1 256 | ax2.plot(y[6, 1:j], y[3, 1:j] * 1E2, '.-', linewidth=2, markersize=8, 257 | label=f.stem.replace('study_coco_', '').replace('yolo', 'YOLO')) 258 | 259 | ax2.plot(1E3 / np.array([209, 140, 97, 58, 35, 18]), [34.6, 40.5, 43.0, 47.5, 49.7, 51.5], 260 | 'k.-', linewidth=2, markersize=8, alpha=.25, label='EfficientDet') 261 | 262 | ax2.grid(alpha=0.2) 263 | ax2.set_yticks(np.arange(20, 60, 5)) 264 | ax2.set_xlim(0, 57) 265 | ax2.set_ylim(30, 55) 266 | ax2.set_xlabel('GPU Speed (ms/img)') 267 | ax2.set_ylabel('COCO AP val') 268 | ax2.legend(loc='lower right') 269 | plt.savefig(str(Path(path).name) + '.png', dpi=300) 270 | 271 | 272 | def plot_labels(labels, names=(), save_dir=Path(''), loggers=None): 273 | # plot dataset labels 274 | print('Plotting labels... ') 275 | c, b = labels[:, 0], labels[:, 1:].transpose() # classes, boxes 276 | nc = int(c.max() + 1) # number of classes 277 | colors = color_list() 278 | x = pd.DataFrame(b.transpose(), columns=['x', 'y', 'width', 'height']) 279 | 280 | # seaborn correlogram 281 | sns.pairplot(x, corner=True, diag_kind='auto', kind='hist', diag_kws=dict(bins=50), plot_kws=dict(pmax=0.9)) 282 | plt.savefig(save_dir / 'labels_correlogram.jpg', dpi=200) 283 | plt.close() 284 | 285 | # matplotlib labels 286 | matplotlib.use('svg') # faster 287 | ax = plt.subplots(2, 2, figsize=(8, 8), tight_layout=True)[1].ravel() 288 | ax[0].hist(c, bins=np.linspace(0, nc, nc + 1) - 0.5, rwidth=0.8) 289 | ax[0].set_ylabel('instances') 290 | if 0 < len(names) < 30: 291 | ax[0].set_xticks(range(len(names))) 292 | ax[0].set_xticklabels(names, rotation=90, fontsize=10) 293 | else: 294 | ax[0].set_xlabel('classes') 295 | sns.histplot(x, x='x', y='y', ax=ax[2], bins=50, pmax=0.9) 296 | sns.histplot(x, x='width', y='height', ax=ax[3], bins=50, pmax=0.9) 297 | 298 | # rectangles 299 | labels[:, 1:3] = 0.5 # center 300 | labels[:, 1:] = xywh2xyxy(labels[:, 1:]) * 2000 301 | img = Image.fromarray(np.ones((2000, 2000, 3), dtype=np.uint8) * 255) 302 | for cls, *box in labels[:1000]: 303 | ImageDraw.Draw(img).rectangle(box, width=1, outline=colors[int(cls) % 10]) # plot 304 | ax[1].imshow(img) 305 | ax[1].axis('off') 306 | 307 | for a in [0, 1, 2, 3]: 308 | for s in ['top', 'right', 'left', 'bottom']: 309 | ax[a].spines[s].set_visible(False) 310 | 311 | plt.savefig(save_dir / 'labels.jpg', dpi=200) 312 | matplotlib.use('Agg') 313 | plt.close() 314 | 315 | # loggers 316 | for k, v in loggers.items() or {}: 317 | if k == 'wandb' and v: 318 | v.log({"Labels": [v.Image(str(x), caption=x.name) for x in save_dir.glob('*labels*.jpg')]}, commit=False) 319 | 320 | 321 | def plot_evolution(yaml_file='data/hyp.finetune.yaml'): # from utils.plots import *; plot_evolution() 322 | # Plot hyperparameter evolution results in evolve.txt 323 | with open(yaml_file) as f: 324 | hyp = yaml.load(f, Loader=yaml.SafeLoader) 325 | x = np.loadtxt('evolve.txt', ndmin=2) 326 | f = fitness(x) 327 | # weights = (f - f.min()) ** 2 # for weighted results 328 | plt.figure(figsize=(10, 12), tight_layout=True) 329 | matplotlib.rc('font', **{'size': 8}) 330 | for i, (k, v) in enumerate(hyp.items()): 331 | y = x[:, i + 7] 332 | # mu = (y * weights).sum() / weights.sum() # best weighted result 333 | mu = y[f.argmax()] # best single result 334 | plt.subplot(6, 5, i + 1) 335 | plt.scatter(y, f, c=hist2d(y, f, 20), cmap='viridis', alpha=.8, edgecolors='none') 336 | plt.plot(mu, f.max(), 'k+', markersize=15) 337 | plt.title('%s = %.3g' % (k, mu), fontdict={'size': 9}) # limit to 40 characters 338 | if i % 5 != 0: 339 | plt.yticks([]) 340 | print('%15s: %.3g' % (k, mu)) 341 | plt.savefig('evolve.png', dpi=200) 342 | print('\nPlot saved as evolve.png') 343 | 344 | 345 | def profile_idetection(start=0, stop=0, labels=(), save_dir=''): 346 | # Plot iDetection '*.txt' per-image logs. from utils.plots import *; profile_idetection() 347 | ax = plt.subplots(2, 4, figsize=(12, 6), tight_layout=True)[1].ravel() 348 | s = ['Images', 'Free Storage (GB)', 'RAM Usage (GB)', 'Battery', 'dt_raw (ms)', 'dt_smooth (ms)', 'real-world FPS'] 349 | files = list(Path(save_dir).glob('frames*.txt')) 350 | for fi, f in enumerate(files): 351 | try: 352 | results = np.loadtxt(f, ndmin=2).T[:, 90:-30] # clip first and last rows 353 | n = results.shape[1] # number of rows 354 | x = np.arange(start, min(stop, n) if stop else n) 355 | results = results[:, x] 356 | t = (results[0] - results[0].min()) # set t0=0s 357 | results[0] = x 358 | for i, a in enumerate(ax): 359 | if i < len(results): 360 | label = labels[fi] if len(labels) else f.stem.replace('frames_', '') 361 | a.plot(t, results[i], marker='.', label=label, linewidth=1, markersize=5) 362 | a.set_title(s[i]) 363 | a.set_xlabel('time (s)') 364 | # if fi == len(files) - 1: 365 | # a.set_ylim(bottom=0) 366 | for side in ['top', 'right']: 367 | a.spines[side].set_visible(False) 368 | else: 369 | a.remove() 370 | except Exception as e: 371 | print('Warning: Plotting error for %s; %s' % (f, e)) 372 | 373 | ax[1].legend() 374 | plt.savefig(Path(save_dir) / 'idetection_profile.png', dpi=200) 375 | 376 | 377 | def plot_results_overlay(start=0, stop=0): # from utils.plots import *; plot_results_overlay() 378 | # Plot training 'results*.txt', overlaying train and val losses 379 | s = ['train', 'train', 'train', 'Precision', 'mAP@0.5', 'val', 'val', 'val', 'Recall', 'mAP@0.5:0.95'] # legends 380 | t = ['Box', 'Objectness', 'Classification', 'P-R', 'mAP-F1'] # titles 381 | for f in sorted(glob.glob('results*.txt') + glob.glob('../../Downloads/results*.txt')): 382 | results = np.loadtxt(f, usecols=[2, 3, 4, 8, 9, 12, 13, 14, 10, 11], ndmin=2).T 383 | n = results.shape[1] # number of rows 384 | x = range(start, min(stop, n) if stop else n) 385 | fig, ax = plt.subplots(1, 5, figsize=(14, 3.5), tight_layout=True) 386 | ax = ax.ravel() 387 | for i in range(5): 388 | for j in [i, i + 5]: 389 | y = results[j, x] 390 | ax[i].plot(x, y, marker='.', label=s[j]) 391 | # y_smooth = butter_lowpass_filtfilt(y) 392 | # ax[i].plot(x, np.gradient(y_smooth), marker='.', label=s[j]) 393 | 394 | ax[i].set_title(t[i]) 395 | ax[i].legend() 396 | ax[i].set_ylabel(f) if i == 0 else None # add filename 397 | fig.savefig(f.replace('.txt', '.png'), dpi=200) 398 | 399 | 400 | def plot_results(start=0, stop=0, bucket='', id=(), labels=(), save_dir=''): 401 | # Plot training 'results*.txt'. from utils.plots import *; plot_results(save_dir='runs/train/exp') 402 | fig, ax = plt.subplots(2, 5, figsize=(12, 6), tight_layout=True) 403 | ax = ax.ravel() 404 | s = ['Box', 'Objectness', 'Classification', 'Precision', 'Recall', 405 | 'val Box', 'val Objectness', 'val Classification', 'mAP@0.5', 'mAP@0.5:0.95'] 406 | if bucket: 407 | # files = ['https://storage.googleapis.com/%s/results%g.txt' % (bucket, x) for x in id] 408 | files = ['results%g.txt' % x for x in id] 409 | c = ('gsutil cp ' + '%s ' * len(files) + '.') % tuple('gs://%s/results%g.txt' % (bucket, x) for x in id) 410 | os.system(c) 411 | else: 412 | files = list(Path(save_dir).glob('results*.txt')) 413 | assert len(files), 'No results.txt files found in %s, nothing to plot.' % os.path.abspath(save_dir) 414 | for fi, f in enumerate(files): 415 | try: 416 | results = np.loadtxt(f, usecols=[2, 3, 4, 8, 9, 12, 13, 14, 10, 11], ndmin=2).T 417 | n = results.shape[1] # number of rows 418 | x = range(start, min(stop, n) if stop else n) 419 | for i in range(10): 420 | y = results[i, x] 421 | if i in [0, 1, 2, 5, 6, 7]: 422 | y[y == 0] = np.nan # don't show zero loss values 423 | # y /= y[0] # normalize 424 | label = labels[fi] if len(labels) else f.stem 425 | ax[i].plot(x, y, marker='.', label=label, linewidth=2, markersize=8) 426 | ax[i].set_title(s[i]) 427 | # if i in [5, 6, 7]: # share train and val loss y axes 428 | # ax[i].get_shared_y_axes().join(ax[i], ax[i - 5]) 429 | except Exception as e: 430 | print('Warning: Plotting error for %s; %s' % (f, e)) 431 | 432 | ax[1].legend() 433 | fig.savefig(Path(save_dir) / 'results.png', dpi=200) 434 | 435 | 436 | def output_to_keypoint(output): 437 | # Convert model output to target format [batch_id, class_id, x, y, w, h, conf] 438 | targets = [] 439 | for i, o in enumerate(output): 440 | kpts = o[:,6:] 441 | o = o[:,:6] 442 | for index, (*box, conf, cls) in enumerate(o.detach().cpu().numpy()): 443 | targets.append([i, cls, *list(*xyxy2xywh(np.array(box)[None])), conf, *list(kpts.detach().cpu().numpy()[index])]) 444 | return np.array(targets) 445 | 446 | 447 | def plot_skeleton_kpts(im, kpts, steps, orig_shape=None): 448 | #Plot the skeleton and keypointsfor coco datatset 449 | palette = np.array([[255, 128, 0], [255, 153, 51], [255, 178, 102], 450 | [230, 230, 0], [255, 153, 255], [153, 204, 255], 451 | [255, 102, 255], [255, 51, 255], [102, 178, 255], 452 | [51, 153, 255], [255, 153, 153], [255, 102, 102], 453 | [255, 51, 51], [153, 255, 153], [102, 255, 102], 454 | [51, 255, 51], [0, 255, 0], [0, 0, 255], [255, 0, 0], 455 | [255, 255, 255]]) 456 | 457 | skeleton = [[16, 14], [14, 12], [17, 15], [15, 13], [12, 13], [6, 12], 458 | [7, 13], [6, 7], [6, 8], [7, 9], [8, 10], [9, 11], [2, 3], 459 | [1, 2], [1, 3], [2, 4], [3, 5], [4, 6], [5, 7]] 460 | 461 | pose_limb_color = palette[[9, 9, 9, 9, 7, 7, 7, 0, 0, 0, 0, 0, 16, 16, 16, 16, 16, 16, 16]] 462 | pose_kpt_color = palette[[16, 16, 16, 16, 16, 0, 0, 0, 0, 0, 0, 9, 9, 9, 9, 9, 9]] 463 | radius = 5 464 | num_kpts = len(kpts) // steps 465 | 466 | for kid in range(num_kpts): 467 | r, g, b = pose_kpt_color[kid] 468 | x_coord, y_coord = kpts[steps * kid], kpts[steps * kid + 1] 469 | if not (x_coord % 640 == 0 or y_coord % 640 == 0): 470 | if steps == 3: 471 | conf = kpts[steps * kid + 2] 472 | if conf < 0.5: 473 | continue 474 | cv2.circle(im, (int(x_coord), int(y_coord)), radius, (int(r), int(g), int(b)), -1) 475 | 476 | for sk_id, sk in enumerate(skeleton): 477 | r, g, b = pose_limb_color[sk_id] 478 | pos1 = (int(kpts[(sk[0]-1)*steps]), int(kpts[(sk[0]-1)*steps+1])) 479 | pos2 = (int(kpts[(sk[1]-1)*steps]), int(kpts[(sk[1]-1)*steps+1])) 480 | if steps == 3: 481 | conf1 = kpts[(sk[0]-1)*steps+2] 482 | conf2 = kpts[(sk[1]-1)*steps+2] 483 | if conf1<0.5 or conf2<0.5: 484 | continue 485 | if pos1[0]%640 == 0 or pos1[1]%640==0 or pos1[0]<0 or pos1[1]<0: 486 | continue 487 | if pos2[0] % 640 == 0 or pos2[1] % 640 == 0 or pos2[0]<0 or pos2[1]<0: 488 | continue 489 | cv2.line(im, pos1, pos2, (int(r), int(g), int(b)), thickness=2) 490 | -------------------------------------------------------------------------------- /utils/torch_utils.py: -------------------------------------------------------------------------------- 1 | # YOLOR PyTorch utils 2 | 3 | import datetime 4 | import logging 5 | import math 6 | import os 7 | import platform 8 | import subprocess 9 | import time 10 | from contextlib import contextmanager 11 | from copy import deepcopy 12 | from pathlib import Path 13 | 14 | import torch 15 | import torch.backends.cudnn as cudnn 16 | import torch.nn as nn 17 | import torch.nn.functional as F 18 | import torchvision 19 | 20 | try: 21 | import thop # for FLOPS computation 22 | except ImportError: 23 | thop = None 24 | logger = logging.getLogger(__name__) 25 | 26 | 27 | @contextmanager 28 | def torch_distributed_zero_first(local_rank: int): 29 | """ 30 | Decorator to make all processes in distributed training wait for each local_master to do something. 31 | """ 32 | if local_rank not in [-1, 0]: 33 | torch.distributed.barrier() 34 | yield 35 | if local_rank == 0: 36 | torch.distributed.barrier() 37 | 38 | 39 | def init_torch_seeds(seed=0): 40 | # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html 41 | torch.manual_seed(seed) 42 | if seed == 0: # slower, more reproducible 43 | cudnn.benchmark, cudnn.deterministic = False, True 44 | else: # faster, less reproducible 45 | cudnn.benchmark, cudnn.deterministic = True, False 46 | 47 | 48 | def date_modified(path=__file__): 49 | # return human-readable file modification date, i.e. '2021-3-26' 50 | t = datetime.datetime.fromtimestamp(Path(path).stat().st_mtime) 51 | return f'{t.year}-{t.month}-{t.day}' 52 | 53 | 54 | def git_describe(path=Path(__file__).parent): # path must be a directory 55 | # return human-readable git description, i.e. v5.0-5-g3e25f1e https://git-scm.com/docs/git-describe 56 | s = f'git -C {path} describe --tags --long --always' 57 | try: 58 | return subprocess.check_output(s, shell=True, stderr=subprocess.STDOUT).decode()[:-1] 59 | except subprocess.CalledProcessError as e: 60 | return '' # not a git repository 61 | 62 | 63 | def select_device(device='', batch_size=None): 64 | # device = 'cpu' or '0' or '0,1,2,3' 65 | s = f'YOLOR 🚀 {git_describe() or date_modified()} torch {torch.__version__} ' # string 66 | cpu = device.lower() == 'cpu' 67 | if cpu: 68 | os.environ['CUDA_VISIBLE_DEVICES'] = '-1' # force torch.cuda.is_available() = False 69 | elif device: # non-cpu device requested 70 | os.environ['CUDA_VISIBLE_DEVICES'] = device # set environment variable 71 | assert torch.cuda.is_available(), f'CUDA unavailable, invalid device {device} requested' # check availability 72 | 73 | cuda = not cpu and torch.cuda.is_available() 74 | if cuda: 75 | n = torch.cuda.device_count() 76 | if n > 1 and batch_size: # check that batch_size is compatible with device_count 77 | assert batch_size % n == 0, f'batch-size {batch_size} not multiple of GPU count {n}' 78 | space = ' ' * len(s) 79 | for i, d in enumerate(device.split(',') if device else range(n)): 80 | p = torch.cuda.get_device_properties(i) 81 | s += f"{'' if i == 0 else space}CUDA:{d} ({p.name}, {p.total_memory / 1024 ** 2}MB)\n" # bytes to MB 82 | else: 83 | s += 'CPU\n' 84 | 85 | logger.info(s.encode().decode('ascii', 'ignore') if platform.system() == 'Windows' else s) # emoji-safe 86 | return torch.device('cuda:0' if cuda else 'cpu') 87 | 88 | 89 | def time_synchronized(): 90 | # pytorch-accurate time 91 | if torch.cuda.is_available(): 92 | torch.cuda.synchronize() 93 | return time.time() 94 | 95 | 96 | def profile(x, ops, n=100, device=None): 97 | # profile a pytorch module or list of modules. Example usage: 98 | # x = torch.randn(16, 3, 640, 640) # input 99 | # m1 = lambda x: x * torch.sigmoid(x) 100 | # m2 = nn.SiLU() 101 | # profile(x, [m1, m2], n=100) # profile speed over 100 iterations 102 | 103 | device = device or torch.device('cuda:0' if torch.cuda.is_available() else 'cpu') 104 | x = x.to(device) 105 | x.requires_grad = True 106 | print(torch.__version__, device.type, torch.cuda.get_device_properties(0) if device.type == 'cuda' else '') 107 | print(f"\n{'Params':>12s}{'GFLOPS':>12s}{'forward (ms)':>16s}{'backward (ms)':>16s}{'input':>24s}{'output':>24s}") 108 | for m in ops if isinstance(ops, list) else [ops]: 109 | m = m.to(device) if hasattr(m, 'to') else m # device 110 | m = m.half() if hasattr(m, 'half') and isinstance(x, torch.Tensor) and x.dtype is torch.float16 else m # type 111 | dtf, dtb, t = 0., 0., [0., 0., 0.] # dt forward, backward 112 | try: 113 | flops = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 # GFLOPS 114 | except: 115 | flops = 0 116 | 117 | for _ in range(n): 118 | t[0] = time_synchronized() 119 | y = m(x) 120 | t[1] = time_synchronized() 121 | try: 122 | _ = y.sum().backward() 123 | t[2] = time_synchronized() 124 | except: # no backward method 125 | t[2] = float('nan') 126 | dtf += (t[1] - t[0]) * 1000 / n # ms per op forward 127 | dtb += (t[2] - t[1]) * 1000 / n # ms per op backward 128 | 129 | s_in = tuple(x.shape) if isinstance(x, torch.Tensor) else 'list' 130 | s_out = tuple(y.shape) if isinstance(y, torch.Tensor) else 'list' 131 | p = sum(list(x.numel() for x in m.parameters())) if isinstance(m, nn.Module) else 0 # parameters 132 | print(f'{p:12}{flops:12.4g}{dtf:16.4g}{dtb:16.4g}{str(s_in):>24s}{str(s_out):>24s}') 133 | 134 | 135 | def is_parallel(model): 136 | return type(model) in (nn.parallel.DataParallel, nn.parallel.DistributedDataParallel) 137 | 138 | 139 | def intersect_dicts(da, db, exclude=()): 140 | # Dictionary intersection of matching keys and shapes, omitting 'exclude' keys, using da values 141 | return {k: v for k, v in da.items() if k in db and not any(x in k for x in exclude) and v.shape == db[k].shape} 142 | 143 | 144 | def initialize_weights(model): 145 | for m in model.modules(): 146 | t = type(m) 147 | if t is nn.Conv2d: 148 | pass # nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') 149 | elif t is nn.BatchNorm2d: 150 | m.eps = 1e-3 151 | m.momentum = 0.03 152 | elif t in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6]: 153 | m.inplace = True 154 | 155 | 156 | def find_modules(model, mclass=nn.Conv2d): 157 | # Finds layer indices matching module class 'mclass' 158 | return [i for i, m in enumerate(model.module_list) if isinstance(m, mclass)] 159 | 160 | 161 | def sparsity(model): 162 | # Return global model sparsity 163 | a, b = 0., 0. 164 | for p in model.parameters(): 165 | a += p.numel() 166 | b += (p == 0).sum() 167 | return b / a 168 | 169 | 170 | def prune(model, amount=0.3): 171 | # Prune model to requested global sparsity 172 | import torch.nn.utils.prune as prune 173 | print('Pruning model... ', end='') 174 | for name, m in model.named_modules(): 175 | if isinstance(m, nn.Conv2d): 176 | prune.l1_unstructured(m, name='weight', amount=amount) # prune 177 | prune.remove(m, 'weight') # make permanent 178 | print(' %.3g global sparsity' % sparsity(model)) 179 | 180 | 181 | def fuse_conv_and_bn(conv, bn): 182 | # Fuse convolution and batchnorm layers https://tehnokv.com/posts/fusing-batchnorm-and-conv/ 183 | fusedconv = nn.Conv2d(conv.in_channels, 184 | conv.out_channels, 185 | kernel_size=conv.kernel_size, 186 | stride=conv.stride, 187 | padding=conv.padding, 188 | groups=conv.groups, 189 | bias=True).requires_grad_(False).to(conv.weight.device) 190 | 191 | # prepare filters 192 | w_conv = conv.weight.clone().view(conv.out_channels, -1) 193 | w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var))) 194 | fusedconv.weight.copy_(torch.mm(w_bn, w_conv).view(fusedconv.weight.shape)) 195 | 196 | # prepare spatial bias 197 | b_conv = torch.zeros(conv.weight.size(0), device=conv.weight.device) if conv.bias is None else conv.bias 198 | b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps)) 199 | fusedconv.bias.copy_(torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn) 200 | 201 | return fusedconv 202 | 203 | 204 | def model_info(model, verbose=False, img_size=640): 205 | # Model information. img_size may be int or list, i.e. img_size=640 or img_size=[640, 320] 206 | n_p = sum(x.numel() for x in model.parameters()) # number parameters 207 | n_g = sum(x.numel() for x in model.parameters() if x.requires_grad) # number gradients 208 | if verbose: 209 | print('%5s %40s %9s %12s %20s %10s %10s' % ('layer', 'name', 'gradient', 'parameters', 'shape', 'mu', 'sigma')) 210 | for i, (name, p) in enumerate(model.named_parameters()): 211 | name = name.replace('module_list.', '') 212 | print('%5g %40s %9s %12g %20s %10.3g %10.3g' % 213 | (i, name, p.requires_grad, p.numel(), list(p.shape), p.mean(), p.std())) 214 | 215 | try: # FLOPS 216 | from thop import profile 217 | stride = max(int(model.stride.max()), 32) if hasattr(model, 'stride') else 32 218 | img = torch.zeros((1, model.yaml.get('ch', 3), stride, stride), device=next(model.parameters()).device) # input 219 | flops = profile(deepcopy(model), inputs=(img,), verbose=False)[0] / 1E9 * 2 # stride GFLOPS 220 | img_size = img_size if isinstance(img_size, list) else [img_size, img_size] # expand if int/float 221 | fs = ', %.1f GFLOPS' % (flops * img_size[0] / stride * img_size[1] / stride) # 640x640 GFLOPS 222 | except (ImportError, Exception): 223 | fs = '' 224 | 225 | logger.info(f"Model Summary: {len(list(model.modules()))} layers, {n_p} parameters, {n_g} gradients{fs}") 226 | 227 | 228 | def load_classifier(name='resnet101', n=2): 229 | # Loads a pretrained model reshaped to n-class output 230 | model = torchvision.models.__dict__[name](pretrained=True) 231 | 232 | # ResNet model properties 233 | # input_size = [3, 224, 224] 234 | # input_space = 'RGB' 235 | # input_range = [0, 1] 236 | # mean = [0.485, 0.456, 0.406] 237 | # std = [0.229, 0.224, 0.225] 238 | 239 | # Reshape output to n classes 240 | filters = model.fc.weight.shape[1] 241 | model.fc.bias = nn.Parameter(torch.zeros(n), requires_grad=True) 242 | model.fc.weight = nn.Parameter(torch.zeros(n, filters), requires_grad=True) 243 | model.fc.out_features = n 244 | return model 245 | 246 | 247 | def scale_img(img, ratio=1.0, same_shape=False, gs=32): # img(16,3,256,416) 248 | # scales img(bs,3,y,x) by ratio constrained to gs-multiple 249 | if ratio == 1.0: 250 | return img 251 | else: 252 | h, w = img.shape[2:] 253 | s = (int(h * ratio), int(w * ratio)) # new size 254 | img = F.interpolate(img, size=s, mode='bilinear', align_corners=False) # resize 255 | if not same_shape: # pad/crop img 256 | h, w = [math.ceil(x * ratio / gs) * gs for x in (h, w)] 257 | return F.pad(img, [0, w - s[1], 0, h - s[0]], value=0.447) # value = imagenet mean 258 | 259 | 260 | def copy_attr(a, b, include=(), exclude=()): 261 | # Copy attributes from b to a, options to only include [...] and to exclude [...] 262 | for k, v in b.__dict__.items(): 263 | if (len(include) and k not in include) or k.startswith('_') or k in exclude: 264 | continue 265 | else: 266 | setattr(a, k, v) 267 | 268 | 269 | class ModelEMA: 270 | """ Model Exponential Moving Average from https://github.com/rwightman/pytorch-image-models 271 | Keep a moving average of everything in the model state_dict (parameters and buffers). 272 | This is intended to allow functionality like 273 | https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage 274 | A smoothed version of the weights is necessary for some training schemes to perform well. 275 | This class is sensitive where it is initialized in the sequence of model init, 276 | GPU assignment and distributed training wrappers. 277 | """ 278 | 279 | def __init__(self, model, decay=0.9999, updates=0): 280 | # Create EMA 281 | self.ema = deepcopy(model.module if is_parallel(model) else model).eval() # FP32 EMA 282 | # if next(model.parameters()).device.type != 'cpu': 283 | # self.ema.half() # FP16 EMA 284 | self.updates = updates # number of EMA updates 285 | self.decay = lambda x: decay * (1 - math.exp(-x / 2000)) # decay exponential ramp (to help early epochs) 286 | for p in self.ema.parameters(): 287 | p.requires_grad_(False) 288 | 289 | def update(self, model): 290 | # Update EMA parameters 291 | with torch.no_grad(): 292 | self.updates += 1 293 | d = self.decay(self.updates) 294 | 295 | msd = model.module.state_dict() if is_parallel(model) else model.state_dict() # model state_dict 296 | for k, v in self.ema.state_dict().items(): 297 | if v.dtype.is_floating_point: 298 | v *= d 299 | v += (1. - d) * msd[k].detach() 300 | 301 | def update_attr(self, model, include=(), exclude=('process_group', 'reducer')): 302 | # Update EMA attributes 303 | copy_attr(self.ema, model, include, exclude) 304 | 305 | 306 | class BatchNormXd(torch.nn.modules.batchnorm._BatchNorm): 307 | def _check_input_dim(self, input): 308 | # The only difference between BatchNorm1d, BatchNorm2d, BatchNorm3d, etc 309 | # is this method that is overwritten by the sub-class 310 | # This original goal of this method was for tensor sanity checks 311 | # If you're ok bypassing those sanity checks (eg. if you trust your inference 312 | # to provide the right dimensional inputs), then you can just use this method 313 | # for easy conversion from SyncBatchNorm 314 | # (unfortunately, SyncBatchNorm does not store the original class - if it did 315 | # we could return the one that was originally created) 316 | return 317 | 318 | def revert_sync_batchnorm(module): 319 | # this is very similar to the function that it is trying to revert: 320 | # https://github.com/pytorch/pytorch/blob/c8b3686a3e4ba63dc59e5dcfe5db3430df256833/torch/nn/modules/batchnorm.py#L679 321 | module_output = module 322 | if isinstance(module, torch.nn.modules.batchnorm.SyncBatchNorm): 323 | new_cls = BatchNormXd 324 | module_output = BatchNormXd(module.num_features, 325 | module.eps, module.momentum, 326 | module.affine, 327 | module.track_running_stats) 328 | if module.affine: 329 | with torch.no_grad(): 330 | module_output.weight = module.weight 331 | module_output.bias = module.bias 332 | module_output.running_mean = module.running_mean 333 | module_output.running_var = module.running_var 334 | module_output.num_batches_tracked = module.num_batches_tracked 335 | if hasattr(module, "qconfig"): 336 | module_output.qconfig = module.qconfig 337 | for name, child in module.named_children(): 338 | module_output.add_module(name, revert_sync_batchnorm(child)) 339 | del module 340 | return module_output 341 | 342 | 343 | class TracedModel(nn.Module): 344 | 345 | def __init__(self, model=None, device=None, img_size=(640,640)): 346 | super(TracedModel, self).__init__() 347 | 348 | print(" Convert model to Traced-model... ") 349 | self.stride = model.stride 350 | self.names = model.names 351 | self.model = model 352 | 353 | self.model = revert_sync_batchnorm(self.model) 354 | self.model.to('cpu') 355 | self.model.eval() 356 | 357 | self.detect_layer = self.model.model[-1] 358 | self.model.traced = True 359 | 360 | rand_example = torch.rand(1, 3, img_size, img_size) 361 | 362 | traced_script_module = torch.jit.trace(self.model, rand_example, strict=False) 363 | #traced_script_module = torch.jit.script(self.model) 364 | traced_script_module.save("traced_model.pt") 365 | print(" traced_script_module saved! ") 366 | self.model = traced_script_module 367 | self.model.to(device) 368 | self.detect_layer.to(device) 369 | print(" model is traced! \n") 370 | 371 | def forward(self, x, augment=False, profile=False): 372 | out = self.model(x) 373 | out = self.detect_layer(out) 374 | return out -------------------------------------------------------------------------------- /utils/wandb_logging/__init__.py: -------------------------------------------------------------------------------- 1 | # init -------------------------------------------------------------------------------- /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.load(f, Loader=yaml.SafeLoader) # 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/coco.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='YOLOR', 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/wandb_logging/wandb_utils.py: -------------------------------------------------------------------------------- 1 | import json 2 | import sys 3 | from pathlib import Path 4 | 5 | import torch 6 | import yaml 7 | from tqdm import tqdm 8 | 9 | sys.path.append(str(Path(__file__).parent.parent.parent)) # add utils/ to path 10 | from utils.datasets import LoadImagesAndLabels 11 | from utils.datasets import img2label_paths 12 | from utils.general import colorstr, xywh2xyxy, check_dataset 13 | 14 | try: 15 | import wandb 16 | from wandb import init, finish 17 | except ImportError: 18 | wandb = None 19 | 20 | WANDB_ARTIFACT_PREFIX = 'wandb-artifact://' 21 | 22 | 23 | def remove_prefix(from_string, prefix=WANDB_ARTIFACT_PREFIX): 24 | return from_string[len(prefix):] 25 | 26 | 27 | def check_wandb_config_file(data_config_file): 28 | wandb_config = '_wandb.'.join(data_config_file.rsplit('.', 1)) # updated data.yaml path 29 | if Path(wandb_config).is_file(): 30 | return wandb_config 31 | return data_config_file 32 | 33 | 34 | def get_run_info(run_path): 35 | run_path = Path(remove_prefix(run_path, WANDB_ARTIFACT_PREFIX)) 36 | run_id = run_path.stem 37 | project = run_path.parent.stem 38 | model_artifact_name = 'run_' + run_id + '_model' 39 | return run_id, project, model_artifact_name 40 | 41 | 42 | def check_wandb_resume(opt): 43 | process_wandb_config_ddp_mode(opt) if opt.global_rank not in [-1, 0] else None 44 | if isinstance(opt.resume, str): 45 | if opt.resume.startswith(WANDB_ARTIFACT_PREFIX): 46 | if opt.global_rank not in [-1, 0]: # For resuming DDP runs 47 | run_id, project, model_artifact_name = get_run_info(opt.resume) 48 | api = wandb.Api() 49 | artifact = api.artifact(project + '/' + model_artifact_name + ':latest') 50 | modeldir = artifact.download() 51 | opt.weights = str(Path(modeldir) / "last.pt") 52 | return True 53 | return None 54 | 55 | 56 | def process_wandb_config_ddp_mode(opt): 57 | with open(opt.data) as f: 58 | data_dict = yaml.load(f, Loader=yaml.SafeLoader) # data dict 59 | train_dir, val_dir = None, None 60 | if isinstance(data_dict['train'], str) and data_dict['train'].startswith(WANDB_ARTIFACT_PREFIX): 61 | api = wandb.Api() 62 | train_artifact = api.artifact(remove_prefix(data_dict['train']) + ':' + opt.artifact_alias) 63 | train_dir = train_artifact.download() 64 | train_path = Path(train_dir) / 'data/images/' 65 | data_dict['train'] = str(train_path) 66 | 67 | if isinstance(data_dict['val'], str) and data_dict['val'].startswith(WANDB_ARTIFACT_PREFIX): 68 | api = wandb.Api() 69 | val_artifact = api.artifact(remove_prefix(data_dict['val']) + ':' + opt.artifact_alias) 70 | val_dir = val_artifact.download() 71 | val_path = Path(val_dir) / 'data/images/' 72 | data_dict['val'] = str(val_path) 73 | if train_dir or val_dir: 74 | ddp_data_path = str(Path(val_dir) / 'wandb_local_data.yaml') 75 | with open(ddp_data_path, 'w') as f: 76 | yaml.dump(data_dict, f) 77 | opt.data = ddp_data_path 78 | 79 | 80 | class WandbLogger(): 81 | def __init__(self, opt, name, run_id, data_dict, job_type='Training'): 82 | # Pre-training routine -- 83 | self.job_type = job_type 84 | self.wandb, self.wandb_run, self.data_dict = wandb, None if not wandb else wandb.run, data_dict 85 | # It's more elegant to stick to 1 wandb.init call, but useful config data is overwritten in the WandbLogger's wandb.init call 86 | if isinstance(opt.resume, str): # checks resume from artifact 87 | if opt.resume.startswith(WANDB_ARTIFACT_PREFIX): 88 | run_id, project, model_artifact_name = get_run_info(opt.resume) 89 | model_artifact_name = WANDB_ARTIFACT_PREFIX + model_artifact_name 90 | assert wandb, 'install wandb to resume wandb runs' 91 | # Resume wandb-artifact:// runs here| workaround for not overwriting wandb.config 92 | self.wandb_run = wandb.init(id=run_id, project=project, resume='allow') 93 | opt.resume = model_artifact_name 94 | elif self.wandb: 95 | self.wandb_run = wandb.init(config=opt, 96 | resume="allow", 97 | project='YOLOR' if opt.project == 'runs/train' else Path(opt.project).stem, 98 | name=name, 99 | job_type=job_type, 100 | id=run_id) if not wandb.run else wandb.run 101 | if self.wandb_run: 102 | if self.job_type == 'Training': 103 | if not opt.resume: 104 | wandb_data_dict = self.check_and_upload_dataset(opt) if opt.upload_dataset else data_dict 105 | # Info useful for resuming from artifacts 106 | self.wandb_run.config.opt = vars(opt) 107 | self.wandb_run.config.data_dict = wandb_data_dict 108 | self.data_dict = self.setup_training(opt, data_dict) 109 | if self.job_type == 'Dataset Creation': 110 | self.data_dict = self.check_and_upload_dataset(opt) 111 | else: 112 | prefix = colorstr('wandb: ') 113 | print(f"{prefix}Install Weights & Biases for YOLOR logging with 'pip install wandb' (recommended)") 114 | 115 | def check_and_upload_dataset(self, opt): 116 | assert wandb, 'Install wandb to upload dataset' 117 | check_dataset(self.data_dict) 118 | config_path = self.log_dataset_artifact(opt.data, 119 | opt.single_cls, 120 | 'YOLOR' if opt.project == 'runs/train' else Path(opt.project).stem) 121 | print("Created dataset config file ", config_path) 122 | with open(config_path) as f: 123 | wandb_data_dict = yaml.load(f, Loader=yaml.SafeLoader) 124 | return wandb_data_dict 125 | 126 | def setup_training(self, opt, data_dict): 127 | self.log_dict, self.current_epoch, self.log_imgs = {}, 0, 16 # Logging Constants 128 | self.bbox_interval = opt.bbox_interval 129 | if isinstance(opt.resume, str): 130 | modeldir, _ = self.download_model_artifact(opt) 131 | if modeldir: 132 | self.weights = Path(modeldir) / "last.pt" 133 | config = self.wandb_run.config 134 | opt.weights, opt.save_period, opt.batch_size, opt.bbox_interval, opt.epochs, opt.hyp = str( 135 | self.weights), config.save_period, config.total_batch_size, config.bbox_interval, config.epochs, \ 136 | config.opt['hyp'] 137 | data_dict = dict(self.wandb_run.config.data_dict) # eliminates the need for config file to resume 138 | if 'val_artifact' not in self.__dict__: # If --upload_dataset is set, use the existing artifact, don't download 139 | self.train_artifact_path, self.train_artifact = self.download_dataset_artifact(data_dict.get('train'), 140 | opt.artifact_alias) 141 | self.val_artifact_path, self.val_artifact = self.download_dataset_artifact(data_dict.get('val'), 142 | opt.artifact_alias) 143 | self.result_artifact, self.result_table, self.val_table, self.weights = None, None, None, None 144 | if self.train_artifact_path is not None: 145 | train_path = Path(self.train_artifact_path) / 'data/images/' 146 | data_dict['train'] = str(train_path) 147 | if self.val_artifact_path is not None: 148 | val_path = Path(self.val_artifact_path) / 'data/images/' 149 | data_dict['val'] = str(val_path) 150 | self.val_table = self.val_artifact.get("val") 151 | self.map_val_table_path() 152 | if self.val_artifact is not None: 153 | self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation") 154 | self.result_table = wandb.Table(["epoch", "id", "prediction", "avg_confidence"]) 155 | if opt.bbox_interval == -1: 156 | self.bbox_interval = opt.bbox_interval = (opt.epochs // 10) if opt.epochs > 10 else 1 157 | return data_dict 158 | 159 | def download_dataset_artifact(self, path, alias): 160 | if isinstance(path, str) and path.startswith(WANDB_ARTIFACT_PREFIX): 161 | dataset_artifact = wandb.use_artifact(remove_prefix(path, WANDB_ARTIFACT_PREFIX) + ":" + alias) 162 | assert dataset_artifact is not None, "'Error: W&B dataset artifact doesn\'t exist'" 163 | datadir = dataset_artifact.download() 164 | return datadir, dataset_artifact 165 | return None, None 166 | 167 | def download_model_artifact(self, opt): 168 | if opt.resume.startswith(WANDB_ARTIFACT_PREFIX): 169 | model_artifact = wandb.use_artifact(remove_prefix(opt.resume, WANDB_ARTIFACT_PREFIX) + ":latest") 170 | assert model_artifact is not None, 'Error: W&B model artifact doesn\'t exist' 171 | modeldir = model_artifact.download() 172 | epochs_trained = model_artifact.metadata.get('epochs_trained') 173 | total_epochs = model_artifact.metadata.get('total_epochs') 174 | assert epochs_trained < total_epochs, 'training to %g epochs is finished, nothing to resume.' % ( 175 | total_epochs) 176 | return modeldir, model_artifact 177 | return None, None 178 | 179 | def log_model(self, path, opt, epoch, fitness_score, best_model=False): 180 | model_artifact = wandb.Artifact('run_' + wandb.run.id + '_model', type='model', metadata={ 181 | 'original_url': str(path), 182 | 'epochs_trained': epoch + 1, 183 | 'save period': opt.save_period, 184 | 'project': opt.project, 185 | 'total_epochs': opt.epochs, 186 | 'fitness_score': fitness_score 187 | }) 188 | model_artifact.add_file(str(path / 'last.pt'), name='last.pt') 189 | wandb.log_artifact(model_artifact, 190 | aliases=['latest', 'epoch ' + str(self.current_epoch), 'best' if best_model else '']) 191 | print("Saving model artifact on epoch ", epoch + 1) 192 | 193 | def log_dataset_artifact(self, data_file, single_cls, project, overwrite_config=False): 194 | with open(data_file) as f: 195 | data = yaml.load(f, Loader=yaml.SafeLoader) # data dict 196 | nc, names = (1, ['item']) if single_cls else (int(data['nc']), data['names']) 197 | names = {k: v for k, v in enumerate(names)} # to index dictionary 198 | self.train_artifact = self.create_dataset_table(LoadImagesAndLabels( 199 | data['train']), names, name='train') if data.get('train') else None 200 | self.val_artifact = self.create_dataset_table(LoadImagesAndLabels( 201 | data['val']), names, name='val') if data.get('val') else None 202 | if data.get('train'): 203 | data['train'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'train') 204 | if data.get('val'): 205 | data['val'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'val') 206 | path = data_file if overwrite_config else '_wandb.'.join(data_file.rsplit('.', 1)) # updated data.yaml path 207 | data.pop('download', None) 208 | with open(path, 'w') as f: 209 | yaml.dump(data, f) 210 | 211 | if self.job_type == 'Training': # builds correct artifact pipeline graph 212 | self.wandb_run.use_artifact(self.val_artifact) 213 | self.wandb_run.use_artifact(self.train_artifact) 214 | self.val_artifact.wait() 215 | self.val_table = self.val_artifact.get('val') 216 | self.map_val_table_path() 217 | else: 218 | self.wandb_run.log_artifact(self.train_artifact) 219 | self.wandb_run.log_artifact(self.val_artifact) 220 | return path 221 | 222 | def map_val_table_path(self): 223 | self.val_table_map = {} 224 | print("Mapping dataset") 225 | for i, data in enumerate(tqdm(self.val_table.data)): 226 | self.val_table_map[data[3]] = data[0] 227 | 228 | def create_dataset_table(self, dataset, class_to_id, name='dataset'): 229 | # TODO: Explore multiprocessing to slpit this loop parallely| This is essential for speeding up the the logging 230 | artifact = wandb.Artifact(name=name, type="dataset") 231 | img_files = tqdm([dataset.path]) if isinstance(dataset.path, str) and Path(dataset.path).is_dir() else None 232 | img_files = tqdm(dataset.img_files) if not img_files else img_files 233 | for img_file in img_files: 234 | if Path(img_file).is_dir(): 235 | artifact.add_dir(img_file, name='data/images') 236 | labels_path = 'labels'.join(dataset.path.rsplit('images', 1)) 237 | artifact.add_dir(labels_path, name='data/labels') 238 | else: 239 | artifact.add_file(img_file, name='data/images/' + Path(img_file).name) 240 | label_file = Path(img2label_paths([img_file])[0]) 241 | artifact.add_file(str(label_file), 242 | name='data/labels/' + label_file.name) if label_file.exists() else None 243 | table = wandb.Table(columns=["id", "train_image", "Classes", "name"]) 244 | class_set = wandb.Classes([{'id': id, 'name': name} for id, name in class_to_id.items()]) 245 | for si, (img, labels, paths, shapes) in enumerate(tqdm(dataset)): 246 | height, width = shapes[0] 247 | labels[:, 2:] = (xywh2xyxy(labels[:, 2:].view(-1, 4))) * torch.Tensor([width, height, width, height]) 248 | box_data, img_classes = [], {} 249 | for cls, *xyxy in labels[:, 1:].tolist(): 250 | cls = int(cls) 251 | box_data.append({"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]}, 252 | "class_id": cls, 253 | "box_caption": "%s" % (class_to_id[cls]), 254 | "scores": {"acc": 1}, 255 | "domain": "pixel"}) 256 | img_classes[cls] = class_to_id[cls] 257 | boxes = {"ground_truth": {"box_data": box_data, "class_labels": class_to_id}} # inference-space 258 | table.add_data(si, wandb.Image(paths, classes=class_set, boxes=boxes), json.dumps(img_classes), 259 | Path(paths).name) 260 | artifact.add(table, name) 261 | return artifact 262 | 263 | def log_training_progress(self, predn, path, names): 264 | if self.val_table and self.result_table: 265 | class_set = wandb.Classes([{'id': id, 'name': name} for id, name in names.items()]) 266 | box_data = [] 267 | total_conf = 0 268 | for *xyxy, conf, cls in predn.tolist(): 269 | if conf >= 0.25: 270 | box_data.append( 271 | {"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]}, 272 | "class_id": int(cls), 273 | "box_caption": "%s %.3f" % (names[cls], conf), 274 | "scores": {"class_score": conf}, 275 | "domain": "pixel"}) 276 | total_conf = total_conf + conf 277 | boxes = {"predictions": {"box_data": box_data, "class_labels": names}} # inference-space 278 | id = self.val_table_map[Path(path).name] 279 | self.result_table.add_data(self.current_epoch, 280 | id, 281 | wandb.Image(self.val_table.data[id][1], boxes=boxes, classes=class_set), 282 | total_conf / max(1, len(box_data)) 283 | ) 284 | 285 | def log(self, log_dict): 286 | if self.wandb_run: 287 | for key, value in log_dict.items(): 288 | self.log_dict[key] = value 289 | 290 | def end_epoch(self, best_result=False): 291 | if self.wandb_run: 292 | wandb.log(self.log_dict) 293 | self.log_dict = {} 294 | if self.result_artifact: 295 | train_results = wandb.JoinedTable(self.val_table, self.result_table, "id") 296 | self.result_artifact.add(train_results, 'result') 297 | wandb.log_artifact(self.result_artifact, aliases=['latest', 'epoch ' + str(self.current_epoch), 298 | ('best' if best_result else '')]) 299 | self.result_table = wandb.Table(["epoch", "id", "prediction", "avg_confidence"]) 300 | self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation") 301 | 302 | def finish_run(self): 303 | if self.wandb_run: 304 | if self.log_dict: 305 | wandb.log(self.log_dict) 306 | wandb.run.finish() 307 | --------------------------------------------------------------------------------