├── README.md ├── Town.gif ├── requirements.txt ├── track.py └── yolov5 ├── Dockerfile ├── LICENSE ├── data ├── coco.yaml ├── coco128.yaml ├── hyp.finetune.yaml ├── hyp.scratch.yaml ├── images │ ├── bus.jpg │ └── zidane.jpg ├── scripts │ ├── get_coco.sh │ └── get_voc.sh └── voc.yaml ├── detect.py ├── hubconf.py └── models ├── common.py ├── experimental.py ├── export.py ├── hub ├── anchors.yaml ├── yolov3-spp.yaml ├── yolov3-tiny.yaml ├── yolov3.yaml ├── yolov5-fpn.yaml ├── yolov5-p2.yaml ├── yolov5-p6.yaml ├── yolov5-p7.yaml └── yolov5-panet.yaml ├── yolo.py ├── yolov5l.yaml ├── yolov5m.yaml ├── yolov5s.yaml └── yolov5x.yaml /README.md: -------------------------------------------------------------------------------- 1 | # Yolov5 + Deep Sort with PyTorch 2 | 3 | [![HitCount](http://hits.dwyl.com/{mikel-brostrom}/{Yolov5_DeepSort_Pytorch}.svg)](http://hits.dwyl.com/{mikel-brostrom}/{Yolov5_DeepSort_Pytorch}) 4 | 5 | 6 | ![](Town.gif) 7 | 8 | ## Introduction 9 | 10 | This repository contains a moded version of PyTorch YOLOv5 (https://github.com/ultralytics/yolov5). It filters out every detection that is not a person. s. The reason behind the fact that it just tracks persons is that the deep association metric is trained on a person ONLY datatset.YOLO（https://github.com/ultralytics） 11 | 12 | ## Description 13 | 14 | The implementation is based on two papers: 15 | 16 | - Simple Online and Realtime Tracking with a Deep Association Metric 17 | https://arxiv.org/abs/1703.07402 18 | - YOLOv4: Optimal Speed and Accuracy of Object Detection 19 | https://arxiv.org/pdf/2004.10934.pdf 20 | 21 | ＃＃ 要求 22 | 23 | Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.7. To install run: 24 | 25 | `pip install -U -r requirements.txt` 26 | 27 | All dependencies are included in the associated docker images. Docker requirements are: 28 | - `nvidia-docker` 29 | - Nvidia Driver Version >= 440.44 30 | 31 | ## Before you run the tracker 32 | 33 | 1. Clone the repository recursively: 34 | 35 | `git clone --recurse-submodules https://github.com/oaqoe-DWQ/Yolov5_DeepSort_Pytorch` 36 | 37 | If you already cloned and forgot to use `--recurse-submodules` you can run `git submodule update --init` 38 | 39 | 2. Github block pushes of files larger than 100 MB (https://help.github.com/en/github/managing-large-files/conditions-for-large-files). Hence you need to download two different weights: the ones for yolo and the ones for deep sort 40 | 41 | - download the yolov5 weight from the latest realease https://github.com/ultralytics/yolov5/releases. Place the downlaoded `.pt` file under `yolov5/weights/` 42 | - download the deep sort weights from https://drive.google.com/drive/folders/1xhG0kRH1EX5B9_Iz8gQJb7UNnn_riXi6. Place ckpt.t7 file under`deep_sort/deep/checkpoint/` 43 | 44 | ## Tracking 45 | 46 | Tracking can be run on most video formats 47 | 48 | ```bash 49 | python3 track.py --source ... 50 | ``` 51 | 52 | - Video: `--source file.mp4` 53 | - Webcam: `--source 0` 54 | - RTSP stream: `--source rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa` 55 | - HTTP stream: `--source http://wmccpinetop.axiscam.net/mjpg/video.mjpg` 56 | 57 | MOT compliant results can be saved to `inference/output` by 58 | 59 | ```bash 60 | python3 track.py --source ... --save-txt 61 | ``` 62 | 63 | ## Other information 64 | 65 | For more detailed information about the algorithms and their corresponding lisences used in this project access their official github implementations. 66 | Thank you for Yolov5 help 67 | 68 | -------------------------------------------------------------------------------- /Town.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/oaqoe-DWQ/Yolov5_DeepSort_Pytorch/2b1a682bf533beb5795bd1debe4469a77b4e2388/Town.gif -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | # pip install -U -r requirements.txt 2 | Cython 3 | matplotlib>=3.2.2 4 | numpy>=1.18.5 5 | opencv-python>=4.1.2 6 | Pillow 7 | PyYAML>=5.3 8 | scipy>=1.4.1 9 | tensorboard>=2.2 10 | torch>=1.7.0 11 | torchvision>=0.8.1 12 | tqdm>=4.41.0 13 | seaborn>=0.11.0 14 | -------------------------------------------------------------------------------- /track.py: -------------------------------------------------------------------------------- 1 | import sys 2 | sys.path.insert(0, './yolov5') 3 | 4 | from yolov5.utils.datasets import LoadImages, LoadStreams 5 | from yolov5.utils.general import check_img_size, non_max_suppression, scale_coords 6 | from yolov5.utils.torch_utils import select_device, time_synchronized 7 | from deep_sort_pytorch.utils.parser import get_config 8 | from deep_sort_pytorch.deep_sort import DeepSort 9 | import argparse 10 | import os 11 | import platform 12 | import shutil 13 | import time 14 | from pathlib import Path 15 | import cv2 16 | import torch 17 | import torch.backends.cudnn as cudnn 18 | 19 | 20 | 21 | palette = (2 ** 11 - 1, 2 ** 15 - 1, 2 ** 20 - 1) 22 | 23 | 24 | def bbox_rel(*xyxy): 25 | """" Calculates the relative bounding box from absolute pixel values. """ 26 | bbox_left = min([xyxy[0].item(), xyxy[2].item()]) 27 | bbox_top = min([xyxy[1].item(), xyxy[3].item()]) 28 | bbox_w = abs(xyxy[0].item() - xyxy[2].item()) 29 | bbox_h = abs(xyxy[1].item() - xyxy[3].item()) 30 | x_c = (bbox_left + bbox_w / 2) 31 | y_c = (bbox_top + bbox_h / 2) 32 | w = bbox_w 33 | h = bbox_h 34 | return x_c, y_c, w, h 35 | 36 | 37 | def compute_color_for_labels(label): 38 | """ 39 | Simple function that adds fixed color depending on the class 40 | """ 41 | color = [int((p * (label ** 2 - label + 1)) % 255) for p in palette] 42 | return tuple(color) 43 | 44 | 45 | def draw_boxes(img, bbox, identities=None, offset=(0, 0)): 46 | for i, box in enumerate(bbox): 47 | x1, y1, x2, y2 = [int(i) for i in box] 48 | x1 += offset[0] 49 | x2 += offset[0] 50 | y1 += offset[1] 51 | y2 += offset[1] 52 | # box text and bar 53 | id = int(identities[i]) if identities is not None else 0 54 | color = compute_color_for_labels(id) 55 | label = '{}{:d}'.format("", id) 56 | t_size = cv2.getTextSize(label, cv2.FONT_HERSHEY_PLAIN, 2, 2)[0] 57 | cv2.rectangle(img, (x1, y1), (x2, y2), color, 3) 58 | cv2.rectangle( 59 | img, (x1, y1), (x1 + t_size[0] + 3, y1 + t_size[1] + 4), color, -1) 60 | cv2.putText(img, label, (x1, y1 + 61 | t_size[1] + 4), cv2.FONT_HERSHEY_PLAIN, 2, [255, 255, 255], 2) 62 | return img 63 | 64 | 65 | def detect(opt, save_img=False): 66 | out, source, weights, view_img, save_txt, imgsz = \ 67 | opt.output, opt.source, opt.weights, opt.view_img, opt.save_txt, opt.img_size 68 | webcam = source == '0' or source.startswith( 69 | 'rtsp') or source.startswith('http') or source.endswith('.txt') 70 | 71 | # initialize deepsort 72 | cfg = get_config() 73 | cfg.merge_from_file(opt.config_deepsort) 74 | deepsort = DeepSort(cfg.DEEPSORT.REID_CKPT, 75 | max_dist=cfg.DEEPSORT.MAX_DIST, min_confidence=cfg.DEEPSORT.MIN_CONFIDENCE, 76 | nms_max_overlap=cfg.DEEPSORT.NMS_MAX_OVERLAP, max_iou_distance=cfg.DEEPSORT.MAX_IOU_DISTANCE, 77 | max_age=cfg.DEEPSORT.MAX_AGE, n_init=cfg.DEEPSORT.N_INIT, nn_budget=cfg.DEEPSORT.NN_BUDGET, 78 | use_cuda=True) 79 | 80 | # Initialize 81 | device = select_device(opt.device) 82 | if os.path.exists(out): 83 | shutil.rmtree(out) # delete output folder 84 | os.makedirs(out) # make new output folder 85 | half = device.type != 'cpu' # half precision only supported on CUDA 86 | 87 | # Load model 88 | model = torch.load(weights, map_location=device)[ 89 | 'model'].float() # load to FP32 90 | model.to(device).eval() 91 | if half: 92 | model.half() # to FP16 93 | 94 | # Set Dataloader 95 | vid_path, vid_writer = None, None 96 | if webcam: 97 | view_img = True 98 | cudnn.benchmark = True # set True to speed up constant image size inference 99 | dataset = LoadStreams(source, img_size=imgsz) 100 | else: 101 | view_img = True 102 | save_img = True 103 | dataset = LoadImages(source, img_size=imgsz) 104 | 105 | # Get names and colors 106 | names = model.module.names if hasattr(model, 'module') else model.names 107 | 108 | # Run inference 109 | t0 = time.time() 110 | img = torch.zeros((1, 3, imgsz, imgsz), device=device) # init img 111 | # run once 112 | _ = model(img.half() if half else img) if device.type != 'cpu' else None 113 | 114 | save_path = str(Path(out)) 115 | txt_path = str(Path(out)) + '/results.txt' 116 | 117 | for frame_idx, (path, img, im0s, vid_cap) in enumerate(dataset): 118 | img = torch.from_numpy(img).to(device) 119 | img = img.half() if half else img.float() # uint8 to fp16/32 120 | img /= 255.0 # 0 - 255 to 0.0 - 1.0 121 | if img.ndimension() == 3: 122 | img = img.unsqueeze(0) 123 | 124 | # Inference 125 | t1 = time_synchronized() 126 | pred = model(img, augment=opt.augment)[0] 127 | 128 | # Apply NMS 129 | pred = non_max_suppression( 130 | pred, opt.conf_thres, opt.iou_thres, classes=opt.classes, agnostic=opt.agnostic_nms) 131 | t2 = time_synchronized() 132 | 133 | # Process detections 134 | for i, det in enumerate(pred): # detections per image 135 | if webcam: # batch_size >= 1 136 | p, s, im0 = path[i], '%g: ' % i, im0s[i].copy() 137 | else: 138 | p, s, im0 = path, '', im0s 139 | 140 | s += '%gx%g ' % img.shape[2:] # print string 141 | save_path = str(Path(out) / Path(p).name) 142 | 143 | if det is not None and len(det): 144 | # Rescale boxes from img_size to im0 size 145 | det[:, :4] = scale_coords( 146 | img.shape[2:], det[:, :4], im0.shape).round() 147 | 148 | # Print results 149 | for c in det[:, -1].unique(): 150 | n = (det[:, -1] == c).sum() # detections per class 151 | s += '%g %ss, ' % (n, names[int(c)]) # add to string 152 | 153 | bbox_xywh = [] 154 | confs = [] 155 | 156 | # Adapt detections to deep sort input format 157 | for *xyxy, conf, cls in det: 158 | x_c, y_c, bbox_w, bbox_h = bbox_rel(*xyxy) 159 | obj = [x_c, y_c, bbox_w, bbox_h] 160 | bbox_xywh.append(obj) 161 | confs.append([conf.item()]) 162 | 163 | xywhs = torch.Tensor(bbox_xywh) 164 | confss = torch.Tensor(confs) 165 | 166 | # Pass detections to deepsort 167 | outputs = deepsort.update(xywhs, confss, im0) 168 | 169 | # draw boxes for visualization 170 | if len(outputs) > 0: 171 | bbox_xyxy = outputs[:, :4] 172 | identities = outputs[:, -1] 173 | draw_boxes(im0, bbox_xyxy, identities) 174 | 175 | # Write MOT compliant results to file 176 | if save_txt and len(outputs) != 0: 177 | for j, output in enumerate(outputs): 178 | bbox_left = output[0] 179 | bbox_top = output[1] 180 | bbox_w = output[2] 181 | bbox_h = output[3] 182 | identity = output[-1] 183 | with open(txt_path, 'a') as f: 184 | f.write(('%g ' * 10 + '\n') % (frame_idx, identity, bbox_left, 185 | bbox_top, bbox_w, bbox_h, -1, -1, -1, -1)) # label format 186 | 187 | else: 188 | deepsort.increment_ages() 189 | 190 | # Print time (inference + NMS) 191 | print('%sDone. (%.3fs)' % (s, t2 - t1)) 192 | 193 | # Stream results 194 | if view_img: 195 | cv2.imshow(p, im0) 196 | if cv2.waitKey(1) == ord('q'): # q to quit 197 | raise StopIteration 198 | 199 | # Save results (image with detections) 200 | if save_img: 201 | print('saving img!') 202 | if dataset.mode == 'images': 203 | cv2.imwrite(save_path, im0) 204 | else: 205 | print('saving video!') 206 | if vid_path != save_path: # new video 207 | vid_path = save_path 208 | if isinstance(vid_writer, cv2.VideoWriter): 209 | vid_writer.release() # release previous video writer 210 | 211 | fps = vid_cap.get(cv2.CAP_PROP_FPS) 212 | w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH)) 213 | h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) 214 | vid_writer = cv2.VideoWriter( 215 | save_path, cv2.VideoWriter_fourcc(*opt.fourcc), fps, (w, h)) 216 | vid_writer.write(im0) 217 | 218 | if save_txt or save_img: 219 | print('Results saved to %s' % os.getcwd() + os.sep + out) 220 | if platform == 'darwin': # MacOS 221 | os.system('open ' + save_path) 222 | 223 | print('Done. (%.3fs)' % (time.time() - t0)) 224 | 225 | 226 | if __name__ == '__main__': 227 | parser = argparse.ArgumentParser() 228 | parser.add_argument('--weights', type=str, 229 | default='yolov5/weights/yolov5s.pt', help='model.pt path') 230 | # file/folder, 0 for webcam 231 | parser.add_argument('--source', type=str, 232 | default='inference/images', help='source') 233 | parser.add_argument('--output', type=str, default='inference/output', 234 | help='output folder') # output folder 235 | parser.add_argument('--img-size', type=int, default=640, 236 | help='inference size (pixels)') 237 | parser.add_argument('--conf-thres', type=float, 238 | default=0.4, help='object confidence threshold') 239 | parser.add_argument('--iou-thres', type=float, 240 | default=0.5, help='IOU threshold for NMS') 241 | parser.add_argument('--fourcc', type=str, default='mp4v', 242 | help='output video codec (verify ffmpeg support)') 243 | parser.add_argument('--device', default='', 244 | help='cuda device, i.e. 0 or 0,1,2,3 or cpu') 245 | parser.add_argument('--view-img', action='store_true', 246 | help='display results') 247 | parser.add_argument('--save-txt', action='store_true', 248 | help='save results to *.txt') 249 | # class 0 is person 250 | parser.add_argument('--classes', nargs='+', type=int, 251 | default=[0], help='filter by class') 252 | parser.add_argument('--agnostic-nms', action='store_true', 253 | help='class-agnostic NMS') 254 | parser.add_argument('--augment', action='store_true', 255 | help='augmented inference') 256 | parser.add_argument("--config_deepsort", type=str, 257 | default="deep_sort_pytorch/configs/deep_sort.yaml") 258 | args = parser.parse_args() 259 | args.img_size = check_img_size(args.img_size) 260 | print(args) 261 | 262 | with torch.no_grad(): 263 | detect(args) 264 | -------------------------------------------------------------------------------- /yolov5/Dockerfile: -------------------------------------------------------------------------------- 1 | # Start FROM Nvidia PyTorch image https://ngc.nvidia.com/catalog/containers/nvidia:pytorch 2 | FROM nvcr.io/nvidia/pytorch:20.12-py3 3 | 4 | # Install linux packages 5 | RUN apt update && apt install -y screen libgl1-mesa-glx 6 | 7 | # Install python dependencies 8 | RUN pip install --upgrade pip 9 | COPY requirements.txt . 10 | RUN pip install -r requirements.txt 11 | RUN pip install gsutil 12 | 13 | # Create working directory 14 | RUN mkdir -p /usr/src/app 15 | WORKDIR /usr/src/app 16 | 17 | # Copy contents 18 | COPY . /usr/src/app 19 | 20 | # Copy weights 21 | #RUN python3 -c "from models import *; \ 22 | #attempt_download('weights/yolov5s.pt'); \ 23 | #attempt_download('weights/yolov5m.pt'); \ 24 | #attempt_download('weights/yolov5l.pt')" 25 | 26 | 27 | # --------------------------------------------------- Extras Below --------------------------------------------------- 28 | 29 | # Build and Push 30 | # t=ultralytics/yolov5:latest && sudo docker build -t $t . && sudo docker push $t 31 | # for v in {300..303}; do t=ultralytics/coco:v$v && sudo docker build -t $t . && sudo docker push $t; done 32 | 33 | # Pull and Run 34 | # t=ultralytics/yolov5:latest && sudo docker pull $t && sudo docker run -it --ipc=host --gpus all $t 35 | 36 | # Pull and Run with local directory access 37 | # t=ultralytics/yolov5:latest && sudo docker pull $t && sudo docker run -it --ipc=host --gpus all -v "$(pwd)"/coco:/usr/src/coco $t 38 | 39 | # Kill all 40 | # sudo docker kill $(sudo docker ps -q) 41 | 42 | # Kill all image-based 43 | # sudo docker kill $(sudo docker ps -a -q --filter ancestor=ultralytics/yolov5:latest) 44 | 45 | # Bash into running container 46 | # sudo docker container exec -it ba65811811ab bash 47 | 48 | # Bash into stopped container 49 | # sudo docker commit 092b16b25c5b usr/resume && sudo docker run -it --gpus all --ipc=host -v "$(pwd)"/coco:/usr/src/coco --entrypoint=sh usr/resume 50 | 51 | # Send weights to GCP 52 | # python -c "from utils.general import *; strip_optimizer('runs/train/exp0_*/weights/best.pt', 'tmp.pt')" && gsutil cp tmp.pt gs://*.pt 53 | 54 | # Clean up 55 | # docker system prune -a --volumes 56 | -------------------------------------------------------------------------------- /yolov5/LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . -------------------------------------------------------------------------------- /yolov5/data/coco.yaml: -------------------------------------------------------------------------------- 1 | # COCO 2017 dataset http://cocodataset.org 2 | # Train command: python train.py --data coco.yaml 3 | # Default dataset location is next to /yolov5: 4 | # /parent_folder 5 | # /coco 6 | # /yolov5 7 | 8 | 9 | # download command/URL (optional) 10 | download: bash data/scripts/get_coco.sh 11 | 12 | # train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/] 13 | train: ../coco/train2017.txt # 118287 images 14 | val: ../coco/val2017.txt # 5000 images 15 | test: ../coco/test-dev2017.txt # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794 16 | 17 | # number of classes 18 | nc: 80 19 | 20 | # class names 21 | names: [ 'person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', 22 | 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', 23 | 'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee', 24 | 'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard', 25 | 'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple', 26 | 'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch', 27 | 'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone', 28 | 'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear', 29 | 'hair drier', 'toothbrush' ] 30 | 31 | # Print classes 32 | # with open('data/coco.yaml') as f: 33 | # d = yaml.load(f, Loader=yaml.FullLoader) # dict 34 | # for i, x in enumerate(d['names']): 35 | # print(i, x) 36 | -------------------------------------------------------------------------------- /yolov5/data/coco128.yaml: -------------------------------------------------------------------------------- 1 | # COCO 2017 dataset http://cocodataset.org - first 128 training images 2 | # Train command: python train.py --data coco128.yaml 3 | # Default dataset location is next to /yolov5: 4 | # /parent_folder 5 | # /coco128 6 | # /yolov5 7 | 8 | 9 | # download command/URL (optional) 10 | download: https://github.com/ultralytics/yolov5/releases/download/v1.0/coco128.zip 11 | 12 | # train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/] 13 | train: ../coco128/images/train2017/ # 128 images 14 | val: ../coco128/images/train2017/ # 128 images 15 | 16 | # number of classes 17 | nc: 80 18 | 19 | # class names 20 | names: [ 'person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', 21 | 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', 22 | 'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee', 23 | 'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard', 24 | 'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple', 25 | 'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch', 26 | 'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone', 27 | 'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear', 28 | 'hair drier', 'toothbrush' ] 29 | -------------------------------------------------------------------------------- /yolov5/data/hyp.finetune.yaml: -------------------------------------------------------------------------------- 1 | # Hyperparameters for VOC finetuning 2 | # python train.py --batch 64 --weights yolov5m.pt --data voc.yaml --img 512 --epochs 50 3 | # See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials 4 | 5 | 6 | # Hyperparameter Evolution Results 7 | # Generations: 306 8 | # P R mAP.5 mAP.5:.95 box obj cls 9 | # Metrics: 0.6 0.936 0.896 0.684 0.0115 0.00805 0.00146 10 | 11 | lr0: 0.0032 12 | lrf: 0.12 13 | momentum: 0.843 14 | weight_decay: 0.00036 15 | warmup_epochs: 2.0 16 | warmup_momentum: 0.5 17 | warmup_bias_lr: 0.05 18 | box: 0.0296 19 | cls: 0.243 20 | cls_pw: 0.631 21 | obj: 0.301 22 | obj_pw: 0.911 23 | iou_t: 0.2 24 | anchor_t: 2.91 25 | # anchors: 3.63 26 | fl_gamma: 0.0 27 | hsv_h: 0.0138 28 | hsv_s: 0.664 29 | hsv_v: 0.464 30 | degrees: 0.373 31 | translate: 0.245 32 | scale: 0.898 33 | shear: 0.602 34 | perspective: 0.0 35 | flipud: 0.00856 36 | fliplr: 0.5 37 | mosaic: 1.0 38 | mixup: 0.243 39 | -------------------------------------------------------------------------------- /yolov5/data/hyp.scratch.yaml: -------------------------------------------------------------------------------- 1 | # Hyperparameters for COCO training from scratch 2 | # python train.py --batch 40 --cfg yolov5m.yaml --weights '' --data coco.yaml --img 640 --epochs 300 3 | # See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials 4 | 5 | 6 | lr0: 0.01 # initial learning rate (SGD=1E-2, Adam=1E-3) 7 | lrf: 0.2 # final OneCycleLR learning rate (lr0 * lrf) 8 | momentum: 0.937 # SGD momentum/Adam beta1 9 | weight_decay: 0.0005 # optimizer weight decay 5e-4 10 | warmup_epochs: 3.0 # warmup epochs (fractions ok) 11 | warmup_momentum: 0.8 # warmup initial momentum 12 | warmup_bias_lr: 0.1 # warmup initial bias lr 13 | box: 0.05 # box loss gain 14 | cls: 0.5 # cls loss gain 15 | cls_pw: 1.0 # cls BCELoss positive_weight 16 | obj: 1.0 # obj loss gain (scale with pixels) 17 | obj_pw: 1.0 # obj BCELoss positive_weight 18 | iou_t: 0.20 # IoU training threshold 19 | anchor_t: 4.0 # anchor-multiple threshold 20 | # anchors: 3 # anchors per output layer (0 to ignore) 21 | fl_gamma: 0.0 # focal loss gamma (efficientDet default gamma=1.5) 22 | hsv_h: 0.015 # image HSV-Hue augmentation (fraction) 23 | hsv_s: 0.7 # image HSV-Saturation augmentation (fraction) 24 | hsv_v: 0.4 # image HSV-Value augmentation (fraction) 25 | degrees: 0.0 # image rotation (+/- deg) 26 | translate: 0.1 # image translation (+/- fraction) 27 | scale: 0.5 # image scale (+/- gain) 28 | shear: 0.0 # image shear (+/- deg) 29 | perspective: 0.0 # image perspective (+/- fraction), range 0-0.001 30 | flipud: 0.0 # image flip up-down (probability) 31 | fliplr: 0.5 # image flip left-right (probability) 32 | mosaic: 1.0 # image mosaic (probability) 33 | mixup: 0.0 # image mixup (probability) 34 | -------------------------------------------------------------------------------- /yolov5/data/images/bus.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/oaqoe-DWQ/Yolov5_DeepSort_Pytorch/2b1a682bf533beb5795bd1debe4469a77b4e2388/yolov5/data/images/bus.jpg -------------------------------------------------------------------------------- /yolov5/data/images/zidane.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/oaqoe-DWQ/Yolov5_DeepSort_Pytorch/2b1a682bf533beb5795bd1debe4469a77b4e2388/yolov5/data/images/zidane.jpg -------------------------------------------------------------------------------- /yolov5/data/scripts/get_coco.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # COCO 2017 dataset http://cocodataset.org 3 | # Download command: bash data/scripts/get_coco.sh 4 | # Train command: python train.py --data coco.yaml 5 | # Default dataset location is next to /yolov5: 6 | # /parent_folder 7 | # /coco 8 | # /yolov5 9 | 10 | # Download/unzip labels 11 | d='../' # unzip directory 12 | url=https://github.com/ultralytics/yolov5/releases/download/v1.0/ 13 | f='coco2017labels.zip' # 68 MB 14 | echo 'Downloading' $url$f ' ...' && curl -L $url$f -o $f && unzip -q $f -d $d && rm $f # download, unzip, remove 15 | 16 | # Download/unzip images 17 | d='../coco/images' # unzip directory 18 | url=http://images.cocodataset.org/zips/ 19 | f1='train2017.zip' # 19G, 118k images 20 | f2='val2017.zip' # 1G, 5k images 21 | f3='test2017.zip' # 7G, 41k images (optional) 22 | for f in $f1 $f2; do 23 | echo 'Downloading' $url$f ' ...' && curl -L $url$f -o $f && unzip -q $f -d $d && rm $f # download, unzip, remove 24 | done 25 | -------------------------------------------------------------------------------- /yolov5/data/scripts/get_voc.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC/ 3 | # Download command: bash data/scripts/get_voc.sh 4 | # Train command: python train.py --data voc.yaml 5 | # Default dataset location is next to /yolov5: 6 | # /parent_folder 7 | # /VOC 8 | # /yolov5 9 | 10 | start=$(date +%s) 11 | mkdir -p ../tmp 12 | cd ../tmp/ 13 | 14 | # Download/unzip images and labels 15 | d='.' # unzip directory 16 | url=https://github.com/ultralytics/yolov5/releases/download/v1.0/ 17 | f1=VOCtrainval_06-Nov-2007.zip # 446MB, 5012 images 18 | f2=VOCtest_06-Nov-2007.zip # 438MB, 4953 images 19 | f3=VOCtrainval_11-May-2012.zip # 1.95GB, 17126 images 20 | for f in $f1 $f2 $f3; do 21 | echo 'Downloading' $url$f ' ...' && curl -L $url$f -o $f && unzip -q $f -d $d && rm $f # download, unzip, remove 22 | done 23 | 24 | end=$(date +%s) 25 | runtime=$((end - start)) 26 | echo "Completed in" $runtime "seconds" 27 | 28 | echo "Splitting dataset..." 29 | python3 - "$@" <train.txt 89 | cat 2007_train.txt 2007_val.txt 2007_test.txt 2012_train.txt 2012_val.txt >train.all.txt 90 | 91 | python3 - "$@" <= 1 84 | p, s, im0, frame = path[i], '%g: ' % i, im0s[i].copy(), dataset.count 85 | else: 86 | p, s, im0, frame = path, '', im0s, getattr(dataset, 'frame', 0) 87 | 88 | p = Path(p) # to Path 89 | save_path = str(save_dir / p.name) # img.jpg 90 | txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}') # img.txt 91 | s += '%gx%g ' % img.shape[2:] # print string 92 | gn = torch.tensor(im0.shape)[[1, 0, 1, 0]] # normalization gain whwh 93 | if len(det): 94 | # Rescale boxes from img_size to im0 size 95 | det[:, :4] = scale_coords(img.shape[2:], det[:, :4], im0.shape).round() 96 | 97 | # Print results 98 | for c in det[:, -1].unique(): 99 | n = (det[:, -1] == c).sum() # detections per class 100 | s += f'{n} {names[int(c)]}s, ' # add to string 101 | 102 | # Write results 103 | for *xyxy, conf, cls in reversed(det): 104 | if save_txt: # Write to file 105 | xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh 106 | line = (cls, *xywh, conf) if opt.save_conf else (cls, *xywh) # label format 107 | with open(txt_path + '.txt', 'a') as f: 108 | f.write(('%g ' * len(line)).rstrip() % line + '\n') 109 | 110 | if save_img or view_img: # Add bbox to image 111 | label = f'{names[int(cls)]} {conf:.2f}' 112 | plot_one_box(xyxy, im0, label=label, color=colors[int(cls)], line_thickness=3) 113 | 114 | # Print time (inference + NMS) 115 | print(f'{s}Done. ({t2 - t1:.3f}s)') 116 | 117 | # Stream results 118 | if view_img: 119 | cv2.imshow(str(p), im0) 120 | 121 | # Save results (image with detections) 122 | if save_img: 123 | if dataset.mode == 'image': 124 | cv2.imwrite(save_path, im0) 125 | else: # 'video' 126 | if vid_path != save_path: # new video 127 | vid_path = save_path 128 | if isinstance(vid_writer, cv2.VideoWriter): 129 | vid_writer.release() # release previous video writer 130 | 131 | fourcc = 'mp4v' # output video codec 132 | fps = vid_cap.get(cv2.CAP_PROP_FPS) 133 | w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH)) 134 | h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) 135 | vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*fourcc), fps, (w, h)) 136 | vid_writer.write(im0) 137 | 138 | if save_txt or save_img: 139 | s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else '' 140 | print(f"Results saved to {save_dir}{s}") 141 | 142 | print(f'Done. ({time.time() - t0:.3f}s)') 143 | 144 | 145 | if __name__ == '__main__': 146 | parser = argparse.ArgumentParser() 147 | parser.add_argument('--weights', nargs='+', type=str, default='yolov5s.pt', help='model.pt path(s)') 148 | parser.add_argument('--source', type=str, default='data/images', help='source') # file/folder, 0 for webcam 149 | parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)') 150 | parser.add_argument('--conf-thres', type=float, default=0.25, help='object confidence threshold') 151 | parser.add_argument('--iou-thres', type=float, default=0.45, help='IOU threshold for NMS') 152 | parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') 153 | parser.add_argument('--view-img', action='store_true', help='display results') 154 | parser.add_argument('--save-txt', action='store_true', help='save results to *.txt') 155 | parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels') 156 | parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3') 157 | parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS') 158 | parser.add_argument('--augment', action='store_true', help='augmented inference') 159 | parser.add_argument('--update', action='store_true', help='update all models') 160 | parser.add_argument('--project', default='runs/detect', help='save results to project/name') 161 | parser.add_argument('--name', default='exp', help='save results to project/name') 162 | parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment') 163 | opt = parser.parse_args() 164 | print(opt) 165 | check_requirements() 166 | 167 | with torch.no_grad(): 168 | if opt.update: # update all models (to fix SourceChangeWarning) 169 | for opt.weights in ['yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt']: 170 | detect() 171 | strip_optimizer(opt.weights) 172 | else: 173 | detect() 174 | -------------------------------------------------------------------------------- /yolov5/hubconf.py: -------------------------------------------------------------------------------- 1 | """File for accessing YOLOv5 via PyTorch Hub https://pytorch.org/hub/ 2 | 3 | Usage: 4 | import torch 5 | model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True, channels=3, classes=80) 6 | """ 7 | 8 | from pathlib import Path 9 | 10 | import torch 11 | 12 | from models.yolo import Model 13 | from utils.general import set_logging 14 | from utils.google_utils import attempt_download 15 | 16 | dependencies = ['torch', 'yaml'] 17 | set_logging() 18 | 19 | 20 | def create(name, pretrained, channels, classes, autoshape): 21 | """Creates a specified YOLOv5 model 22 | 23 | Arguments: 24 | name (str): name of model, i.e. 'yolov5s' 25 | pretrained (bool): load pretrained weights into the model 26 | channels (int): number of input channels 27 | classes (int): number of model classes 28 | 29 | Returns: 30 | pytorch model 31 | """ 32 | config = Path(__file__).parent / 'models' / f'{name}.yaml' # model.yaml path 33 | try: 34 | model = Model(config, channels, classes) 35 | if pretrained: 36 | fname = f'{name}.pt' # checkpoint filename 37 | attempt_download(fname) # download if not found locally 38 | ckpt = torch.load(fname, map_location=torch.device('cpu')) # load 39 | state_dict = ckpt['model'].float().state_dict() # to FP32 40 | state_dict = {k: v for k, v in state_dict.items() if model.state_dict()[k].shape == v.shape} # filter 41 | model.load_state_dict(state_dict, strict=False) # load 42 | if len(ckpt['model'].names) == classes: 43 | model.names = ckpt['model'].names # set class names attribute 44 | if autoshape: 45 | model = model.autoshape() # for file/URI/PIL/cv2/np inputs and NMS 46 | return model 47 | 48 | except Exception as e: 49 | help_url = 'https://github.com/ultralytics/yolov5/issues/36' 50 | s = 'Cache maybe be out of date, try force_reload=True. See %s for help.' % help_url 51 | raise Exception(s) from e 52 | 53 | 54 | def yolov5s(pretrained=False, channels=3, classes=80, autoshape=True): 55 | """YOLOv5-small model from https://github.com/ultralytics/yolov5 56 | 57 | Arguments: 58 | pretrained (bool): load pretrained weights into the model, default=False 59 | channels (int): number of input channels, default=3 60 | classes (int): number of model classes, default=80 61 | 62 | Returns: 63 | pytorch model 64 | """ 65 | return create('yolov5s', pretrained, channels, classes, autoshape) 66 | 67 | 68 | def yolov5m(pretrained=False, channels=3, classes=80, autoshape=True): 69 | """YOLOv5-medium model from https://github.com/ultralytics/yolov5 70 | 71 | Arguments: 72 | pretrained (bool): load pretrained weights into the model, default=False 73 | channels (int): number of input channels, default=3 74 | classes (int): number of model classes, default=80 75 | 76 | Returns: 77 | pytorch model 78 | """ 79 | return create('yolov5m', pretrained, channels, classes, autoshape) 80 | 81 | 82 | def yolov5l(pretrained=False, channels=3, classes=80, autoshape=True): 83 | """YOLOv5-large model from https://github.com/ultralytics/yolov5 84 | 85 | Arguments: 86 | pretrained (bool): load pretrained weights into the model, default=False 87 | channels (int): number of input channels, default=3 88 | classes (int): number of model classes, default=80 89 | 90 | Returns: 91 | pytorch model 92 | """ 93 | return create('yolov5l', pretrained, channels, classes, autoshape) 94 | 95 | 96 | def yolov5x(pretrained=False, channels=3, classes=80, autoshape=True): 97 | """YOLOv5-xlarge model from https://github.com/ultralytics/yolov5 98 | 99 | Arguments: 100 | pretrained (bool): load pretrained weights into the model, default=False 101 | channels (int): number of input channels, default=3 102 | classes (int): number of model classes, default=80 103 | 104 | Returns: 105 | pytorch model 106 | """ 107 | return create('yolov5x', pretrained, channels, classes, autoshape) 108 | 109 | 110 | def custom(path_or_model='path/to/model.pt', autoshape=True): 111 | """YOLOv5-custom model from https://github.com/ultralytics/yolov5 112 | 113 | Arguments (3 options): 114 | path_or_model (str): 'path/to/model.pt' 115 | path_or_model (dict): torch.load('path/to/model.pt') 116 | path_or_model (nn.Module): torch.load('path/to/model.pt')['model'] 117 | 118 | Returns: 119 | pytorch model 120 | """ 121 | model = torch.load(path_or_model) if isinstance(path_or_model, str) else path_or_model # load checkpoint 122 | if isinstance(model, dict): 123 | model = model['model'] # load model 124 | 125 | hub_model = Model(model.yaml).to(next(model.parameters()).device) # create 126 | hub_model.load_state_dict(model.float().state_dict()) # load state_dict 127 | hub_model.names = model.names # class names 128 | return hub_model.autoshape() if autoshape else hub_model 129 | 130 | 131 | if __name__ == '__main__': 132 | model = create(name='yolov5s', pretrained=True, channels=3, classes=80, autoshape=True) # pretrained example 133 | # model = custom(path_or_model='path/to/model.pt') # custom example 134 | 135 | # Verify inference 136 | from PIL import Image 137 | 138 | imgs = [Image.open(x) for x in Path('data/images').glob('*.jpg')] 139 | results = model(imgs) 140 | results.show() 141 | results.print() 142 | -------------------------------------------------------------------------------- /yolov5/models/common.py: -------------------------------------------------------------------------------- 1 | # This file contains modules common to various models 2 | 3 | import math 4 | import numpy as np 5 | import requests 6 | import torch 7 | import torch.nn as nn 8 | from PIL import Image, ImageDraw 9 | 10 | from utils.datasets import letterbox 11 | from utils.general import non_max_suppression, make_divisible, scale_coords, xyxy2xywh 12 | from utils.plots import color_list 13 | 14 | 15 | def autopad(k, p=None): # kernel, padding 16 | # Pad to 'same' 17 | if p is None: 18 | p = k // 2 if isinstance(k, int) else [x // 2 for x in k] # auto-pad 19 | return p 20 | 21 | 22 | def DWConv(c1, c2, k=1, s=1, act=True): 23 | # Depthwise convolution 24 | return Conv(c1, c2, k, s, g=math.gcd(c1, c2), act=act) 25 | 26 | 27 | class Conv(nn.Module): 28 | # Standard convolution 29 | def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups 30 | super(Conv, self).__init__() 31 | self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False) 32 | self.bn = nn.BatchNorm2d(c2) 33 | self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity()) 34 | 35 | def forward(self, x): 36 | return self.act(self.bn(self.conv(x))) 37 | 38 | def fuseforward(self, x): 39 | return self.act(self.conv(x)) 40 | 41 | 42 | class Bottleneck(nn.Module): 43 | # Standard bottleneck 44 | def __init__(self, c1, c2, shortcut=True, g=1, e=0.5): # ch_in, ch_out, shortcut, groups, expansion 45 | super(Bottleneck, self).__init__() 46 | c_ = int(c2 * e) # hidden channels 47 | self.cv1 = Conv(c1, c_, 1, 1) 48 | self.cv2 = Conv(c_, c2, 3, 1, g=g) 49 | self.add = shortcut and c1 == c2 50 | 51 | def forward(self, x): 52 | return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x)) 53 | 54 | 55 | class BottleneckCSP(nn.Module): 56 | # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks 57 | def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion 58 | super(BottleneckCSP, self).__init__() 59 | c_ = int(c2 * e) # hidden channels 60 | self.cv1 = Conv(c1, c_, 1, 1) 61 | self.cv2 = nn.Conv2d(c1, c_, 1, 1, bias=False) 62 | self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False) 63 | self.cv4 = Conv(2 * c_, c2, 1, 1) 64 | self.bn = nn.BatchNorm2d(2 * c_) # applied to cat(cv2, cv3) 65 | self.act = nn.LeakyReLU(0.1, inplace=True) 66 | self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)]) 67 | 68 | def forward(self, x): 69 | y1 = self.cv3(self.m(self.cv1(x))) 70 | y2 = self.cv2(x) 71 | return self.cv4(self.act(self.bn(torch.cat((y1, y2), dim=1)))) 72 | 73 | 74 | class C3(nn.Module): 75 | # CSP Bottleneck with 3 convolutions 76 | def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion 77 | super(C3, self).__init__() 78 | c_ = int(c2 * e) # hidden channels 79 | self.cv1 = Conv(c1, c_, 1, 1) 80 | self.cv2 = Conv(c1, c_, 1, 1) 81 | self.cv3 = Conv(2 * c_, c2, 1) # act=FReLU(c2) 82 | self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)]) 83 | # self.m = nn.Sequential(*[CrossConv(c_, c_, 3, 1, g, 1.0, shortcut) for _ in range(n)]) 84 | 85 | def forward(self, x): 86 | return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1)) 87 | 88 | 89 | class SPP(nn.Module): 90 | # Spatial pyramid pooling layer used in YOLOv3-SPP 91 | def __init__(self, c1, c2, k=(5, 9, 13)): 92 | super(SPP, self).__init__() 93 | c_ = c1 // 2 # hidden channels 94 | self.cv1 = Conv(c1, c_, 1, 1) 95 | self.cv2 = Conv(c_ * (len(k) + 1), c2, 1, 1) 96 | self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k]) 97 | 98 | def forward(self, x): 99 | x = self.cv1(x) 100 | return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1)) 101 | 102 | 103 | class Focus(nn.Module): 104 | # Focus wh information into c-space 105 | def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups 106 | super(Focus, self).__init__() 107 | self.conv = Conv(c1 * 4, c2, k, s, p, g, act) 108 | # self.contract = Contract(gain=2) 109 | 110 | def forward(self, x): # x(b,c,w,h) -> y(b,4c,w/2,h/2) 111 | return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1)) 112 | # return self.conv(self.contract(x)) 113 | 114 | 115 | class Contract(nn.Module): 116 | # Contract width-height into channels, i.e. x(1,64,80,80) to x(1,256,40,40) 117 | def __init__(self, gain=2): 118 | super().__init__() 119 | self.gain = gain 120 | 121 | def forward(self, x): 122 | N, C, H, W = x.size() # assert (H / s == 0) and (W / s == 0), 'Indivisible gain' 123 | s = self.gain 124 | x = x.view(N, C, H // s, s, W // s, s) # x(1,64,40,2,40,2) 125 | x = x.permute(0, 3, 5, 1, 2, 4).contiguous() # x(1,2,2,64,40,40) 126 | return x.view(N, C * s * s, H // s, W // s) # x(1,256,40,40) 127 | 128 | 129 | class Expand(nn.Module): 130 | # Expand channels into width-height, i.e. x(1,64,80,80) to x(1,16,160,160) 131 | def __init__(self, gain=2): 132 | super().__init__() 133 | self.gain = gain 134 | 135 | def forward(self, x): 136 | N, C, H, W = x.size() # assert C / s ** 2 == 0, 'Indivisible gain' 137 | s = self.gain 138 | x = x.view(N, s, s, C // s ** 2, H, W) # x(1,2,2,16,80,80) 139 | x = x.permute(0, 3, 4, 1, 5, 2).contiguous() # x(1,16,80,2,80,2) 140 | return x.view(N, C // s ** 2, H * s, W * s) # x(1,16,160,160) 141 | 142 | 143 | class Concat(nn.Module): 144 | # Concatenate a list of tensors along dimension 145 | def __init__(self, dimension=1): 146 | super(Concat, self).__init__() 147 | self.d = dimension 148 | 149 | def forward(self, x): 150 | return torch.cat(x, self.d) 151 | 152 | 153 | class NMS(nn.Module): 154 | # Non-Maximum Suppression (NMS) module 155 | conf = 0.25 # confidence threshold 156 | iou = 0.45 # IoU threshold 157 | classes = None # (optional list) filter by class 158 | 159 | def __init__(self): 160 | super(NMS, self).__init__() 161 | 162 | def forward(self, x): 163 | return non_max_suppression(x[0], conf_thres=self.conf, iou_thres=self.iou, classes=self.classes) 164 | 165 | 166 | class autoShape(nn.Module): 167 | # input-robust model wrapper for passing cv2/np/PIL/torch inputs. Includes preprocessing, inference and NMS 168 | img_size = 640 # inference size (pixels) 169 | conf = 0.25 # NMS confidence threshold 170 | iou = 0.45 # NMS IoU threshold 171 | classes = None # (optional list) filter by class 172 | 173 | def __init__(self, model): 174 | super(autoShape, self).__init__() 175 | self.model = model.eval() 176 | 177 | def autoshape(self): 178 | print('autoShape already enabled, skipping... ') # model already converted to model.autoshape() 179 | return self 180 | 181 | def forward(self, imgs, size=640, augment=False, profile=False): 182 | # Inference from various sources. For height=720, width=1280, RGB images example inputs are: 183 | # filename: imgs = 'data/samples/zidane.jpg' 184 | # URI: = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/zidane.jpg' 185 | # OpenCV: = cv2.imread('image.jpg')[:,:,::-1] # HWC BGR to RGB x(720,1280,3) 186 | # PIL: = Image.open('image.jpg') # HWC x(720,1280,3) 187 | # numpy: = np.zeros((720,1280,3)) # HWC 188 | # torch: = torch.zeros(16,3,720,1280) # BCHW 189 | # multiple: = [Image.open('image1.jpg'), Image.open('image2.jpg'), ...] # list of images 190 | 191 | p = next(self.model.parameters()) # for device and type 192 | if isinstance(imgs, torch.Tensor): # torch 193 | return self.model(imgs.to(p.device).type_as(p), augment, profile) # inference 194 | 195 | # Pre-process 196 | n, imgs = (len(imgs), imgs) if isinstance(imgs, list) else (1, [imgs]) # number of images, list of images 197 | shape0, shape1 = [], [] # image and inference shapes 198 | for i, im in enumerate(imgs): 199 | if isinstance(im, str): # filename or uri 200 | im = Image.open(requests.get(im, stream=True).raw if im.startswith('http') else im) # open 201 | im = np.array(im) # to numpy 202 | if im.shape[0] < 5: # image in CHW 203 | im = im.transpose((1, 2, 0)) # reverse dataloader .transpose(2, 0, 1) 204 | im = im[:, :, :3] if im.ndim == 3 else np.tile(im[:, :, None], 3) # enforce 3ch input 205 | s = im.shape[:2] # HWC 206 | shape0.append(s) # image shape 207 | g = (size / max(s)) # gain 208 | shape1.append([y * g for y in s]) 209 | imgs[i] = im # update 210 | shape1 = [make_divisible(x, int(self.stride.max())) for x in np.stack(shape1, 0).max(0)] # inference shape 211 | x = [letterbox(im, new_shape=shape1, auto=False)[0] for im in imgs] # pad 212 | x = np.stack(x, 0) if n > 1 else x[0][None] # stack 213 | x = np.ascontiguousarray(x.transpose((0, 3, 1, 2))) # BHWC to BCHW 214 | x = torch.from_numpy(x).to(p.device).type_as(p) / 255. # uint8 to fp16/32 215 | 216 | # Inference 217 | with torch.no_grad(): 218 | y = self.model(x, augment, profile)[0] # forward 219 | y = non_max_suppression(y, conf_thres=self.conf, iou_thres=self.iou, classes=self.classes) # NMS 220 | 221 | # Post-process 222 | for i in range(n): 223 | scale_coords(shape1, y[i][:, :4], shape0[i]) 224 | 225 | return Detections(imgs, y, self.names) 226 | 227 | 228 | class Detections: 229 | # detections class for YOLOv5 inference results 230 | def __init__(self, imgs, pred, names=None): 231 | super(Detections, self).__init__() 232 | d = pred[0].device # device 233 | gn = [torch.tensor([*[im.shape[i] for i in [1, 0, 1, 0]], 1., 1.], device=d) for im in imgs] # normalizations 234 | self.imgs = imgs # list of images as numpy arrays 235 | self.pred = pred # list of tensors pred[0] = (xyxy, conf, cls) 236 | self.names = names # class names 237 | self.xyxy = pred # xyxy pixels 238 | self.xywh = [xyxy2xywh(x) for x in pred] # xywh pixels 239 | self.xyxyn = [x / g for x, g in zip(self.xyxy, gn)] # xyxy normalized 240 | self.xywhn = [x / g for x, g in zip(self.xywh, gn)] # xywh normalized 241 | self.n = len(self.pred) 242 | 243 | def display(self, pprint=False, show=False, save=False): 244 | colors = color_list() 245 | for i, (img, pred) in enumerate(zip(self.imgs, self.pred)): 246 | str = f'Image {i + 1}/{len(self.pred)}: {img.shape[0]}x{img.shape[1]} ' 247 | if pred is not None: 248 | for c in pred[:, -1].unique(): 249 | n = (pred[:, -1] == c).sum() # detections per class 250 | str += f'{n} {self.names[int(c)]}s, ' # add to string 251 | if show or save: 252 | img = Image.fromarray(img.astype(np.uint8)) if isinstance(img, np.ndarray) else img # from np 253 | for *box, conf, cls in pred: # xyxy, confidence, class 254 | # str += '%s %.2f, ' % (names[int(cls)], conf) # label 255 | ImageDraw.Draw(img).rectangle(box, width=4, outline=colors[int(cls) % 10]) # plot 256 | if save: 257 | f = f'results{i}.jpg' 258 | str += f"saved to '{f}'" 259 | img.save(f) # save 260 | if show: 261 | img.show(f'Image {i}') # show 262 | if pprint: 263 | print(str) 264 | 265 | def print(self): 266 | self.display(pprint=True) # print results 267 | 268 | def show(self): 269 | self.display(show=True) # show results 270 | 271 | def save(self): 272 | self.display(save=True) # save results 273 | 274 | def __len__(self): 275 | return self.n 276 | 277 | def tolist(self): 278 | # return a list of Detections objects, i.e. 'for result in results.tolist():' 279 | x = [Detections([self.imgs[i]], [self.pred[i]], self.names) for i in range(self.n)] 280 | for d in x: 281 | for k in ['imgs', 'pred', 'xyxy', 'xyxyn', 'xywh', 'xywhn']: 282 | setattr(d, k, getattr(d, k)[0]) # pop out of list 283 | return x 284 | 285 | 286 | class Classify(nn.Module): 287 | # Classification head, i.e. x(b,c1,20,20) to x(b,c2) 288 | def __init__(self, c1, c2, k=1, s=1, p=None, g=1): # ch_in, ch_out, kernel, stride, padding, groups 289 | super(Classify, self).__init__() 290 | self.aap = nn.AdaptiveAvgPool2d(1) # to x(b,c1,1,1) 291 | self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g) # to x(b,c2,1,1) 292 | self.flat = nn.Flatten() 293 | 294 | def forward(self, x): 295 | z = torch.cat([self.aap(y) for y in (x if isinstance(x, list) else [x])], 1) # cat if list 296 | return self.flat(self.conv(z)) # flatten to x(b,c2) 297 | -------------------------------------------------------------------------------- /yolov5/models/experimental.py: -------------------------------------------------------------------------------- 1 | # This file contains experimental modules 2 | 3 | import numpy as np 4 | import torch 5 | import torch.nn as nn 6 | 7 | from models.common import Conv, DWConv 8 | from utils.google_utils import attempt_download 9 | 10 | 11 | class CrossConv(nn.Module): 12 | # Cross Convolution Downsample 13 | def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False): 14 | # ch_in, ch_out, kernel, stride, groups, expansion, shortcut 15 | super(CrossConv, self).__init__() 16 | c_ = int(c2 * e) # hidden channels 17 | self.cv1 = Conv(c1, c_, (1, k), (1, s)) 18 | self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g) 19 | self.add = shortcut and c1 == c2 20 | 21 | def forward(self, x): 22 | return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x)) 23 | 24 | 25 | class Sum(nn.Module): 26 | # Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070 27 | def __init__(self, n, weight=False): # n: number of inputs 28 | super(Sum, self).__init__() 29 | self.weight = weight # apply weights boolean 30 | self.iter = range(n - 1) # iter object 31 | if weight: 32 | self.w = nn.Parameter(-torch.arange(1., n) / 2, requires_grad=True) # layer weights 33 | 34 | def forward(self, x): 35 | y = x[0] # no weight 36 | if self.weight: 37 | w = torch.sigmoid(self.w) * 2 38 | for i in self.iter: 39 | y = y + x[i + 1] * w[i] 40 | else: 41 | for i in self.iter: 42 | y = y + x[i + 1] 43 | return y 44 | 45 | 46 | class GhostConv(nn.Module): 47 | # Ghost Convolution https://github.com/huawei-noah/ghostnet 48 | def __init__(self, c1, c2, k=1, s=1, g=1, act=True): # ch_in, ch_out, kernel, stride, groups 49 | super(GhostConv, self).__init__() 50 | c_ = c2 // 2 # hidden channels 51 | self.cv1 = Conv(c1, c_, k, s, None, g, act) 52 | self.cv2 = Conv(c_, c_, 5, 1, None, c_, act) 53 | 54 | def forward(self, x): 55 | y = self.cv1(x) 56 | return torch.cat([y, self.cv2(y)], 1) 57 | 58 | 59 | class GhostBottleneck(nn.Module): 60 | # Ghost Bottleneck https://github.com/huawei-noah/ghostnet 61 | def __init__(self, c1, c2, k, s): 62 | super(GhostBottleneck, self).__init__() 63 | c_ = c2 // 2 64 | self.conv = nn.Sequential(GhostConv(c1, c_, 1, 1), # pw 65 | DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(), # dw 66 | GhostConv(c_, c2, 1, 1, act=False)) # pw-linear 67 | self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False), 68 | Conv(c1, c2, 1, 1, act=False)) if s == 2 else nn.Identity() 69 | 70 | def forward(self, x): 71 | return self.conv(x) + self.shortcut(x) 72 | 73 | 74 | class MixConv2d(nn.Module): 75 | # Mixed Depthwise Conv https://arxiv.org/abs/1907.09595 76 | def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True): 77 | super(MixConv2d, self).__init__() 78 | groups = len(k) 79 | if equal_ch: # equal c_ per group 80 | i = torch.linspace(0, groups - 1E-6, c2).floor() # c2 indices 81 | c_ = [(i == g).sum() for g in range(groups)] # intermediate channels 82 | else: # equal weight.numel() per group 83 | b = [c2] + [0] * groups 84 | a = np.eye(groups + 1, groups, k=-1) 85 | a -= np.roll(a, 1, axis=1) 86 | a *= np.array(k) ** 2 87 | a[0] = 1 88 | c_ = np.linalg.lstsq(a, b, rcond=None)[0].round() # solve for equal weight indices, ax = b 89 | 90 | self.m = nn.ModuleList([nn.Conv2d(c1, int(c_[g]), k[g], s, k[g] // 2, bias=False) for g in range(groups)]) 91 | self.bn = nn.BatchNorm2d(c2) 92 | self.act = nn.LeakyReLU(0.1, inplace=True) 93 | 94 | def forward(self, x): 95 | return x + self.act(self.bn(torch.cat([m(x) for m in self.m], 1))) 96 | 97 | 98 | class Ensemble(nn.ModuleList): 99 | # Ensemble of models 100 | def __init__(self): 101 | super(Ensemble, self).__init__() 102 | 103 | def forward(self, x, augment=False): 104 | y = [] 105 | for module in self: 106 | y.append(module(x, augment)[0]) 107 | # y = torch.stack(y).max(0)[0] # max ensemble 108 | # y = torch.stack(y).mean(0) # mean ensemble 109 | y = torch.cat(y, 1) # nms ensemble 110 | return y, None # inference, train output 111 | 112 | 113 | def attempt_load(weights, map_location=None): 114 | # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a 115 | model = Ensemble() 116 | for w in weights if isinstance(weights, list) else [weights]: 117 | attempt_download(w) 118 | model.append(torch.load(w, map_location=map_location)['model'].float().fuse().eval()) # load FP32 model 119 | 120 | # Compatibility updates 121 | for m in model.modules(): 122 | if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6]: 123 | m.inplace = True # pytorch 1.7.0 compatibility 124 | elif type(m) is Conv: 125 | m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility 126 | 127 | if len(model) == 1: 128 | return model[-1] # return model 129 | else: 130 | print('Ensemble created with %s\n' % weights) 131 | for k in ['names', 'stride']: 132 | setattr(model, k, getattr(model[-1], k)) 133 | return model # return ensemble 134 | -------------------------------------------------------------------------------- /yolov5/models/export.py: -------------------------------------------------------------------------------- 1 | """Exports a YOLOv5 *.pt model to ONNX and TorchScript formats 2 | 3 | Usage: 4 | $ export PYTHONPATH="$PWD" && python models/export.py --weights ./weights/yolov5s.pt --img 640 --batch 1 5 | """ 6 | 7 | import argparse 8 | import sys 9 | import time 10 | 11 | sys.path.append('./') # to run '$ python *.py' files in subdirectories 12 | 13 | import torch 14 | import torch.nn as nn 15 | 16 | import models 17 | from models.experimental import attempt_load 18 | from utils.activations import Hardswish, SiLU 19 | from utils.general import set_logging, check_img_size 20 | 21 | if __name__ == '__main__': 22 | parser = argparse.ArgumentParser() 23 | parser.add_argument('--weights', type=str, default='./yolov5s.pt', help='weights path') # from yolov5/models/ 24 | parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='image size') # height, width 25 | parser.add_argument('--batch-size', type=int, default=1, help='batch size') 26 | opt = parser.parse_args() 27 | opt.img_size *= 2 if len(opt.img_size) == 1 else 1 # expand 28 | print(opt) 29 | set_logging() 30 | t = time.time() 31 | 32 | # Load PyTorch model 33 | model = attempt_load(opt.weights, map_location=torch.device('cpu')) # load FP32 model 34 | labels = model.names 35 | 36 | # Checks 37 | gs = int(max(model.stride)) # grid size (max stride) 38 | opt.img_size = [check_img_size(x, gs) for x in opt.img_size] # verify img_size are gs-multiples 39 | 40 | # Input 41 | img = torch.zeros(opt.batch_size, 3, *opt.img_size) # image size(1,3,320,192) iDetection 42 | 43 | # Update model 44 | for k, m in model.named_modules(): 45 | m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility 46 | if isinstance(m, models.common.Conv): # assign export-friendly activations 47 | if isinstance(m.act, nn.Hardswish): 48 | m.act = Hardswish() 49 | elif isinstance(m.act, nn.SiLU): 50 | m.act = SiLU() 51 | # elif isinstance(m, models.yolo.Detect): 52 | # m.forward = m.forward_export # assign forward (optional) 53 | model.model[-1].export = True # set Detect() layer export=True 54 | y = model(img) # dry run 55 | 56 | # TorchScript export 57 | try: 58 | print('\nStarting TorchScript export with torch %s...' % torch.__version__) 59 | f = opt.weights.replace('.pt', '.torchscript.pt') # filename 60 | ts = torch.jit.trace(model, img) 61 | ts.save(f) 62 | print('TorchScript export success, saved as %s' % f) 63 | except Exception as e: 64 | print('TorchScript export failure: %s' % e) 65 | 66 | # ONNX export 67 | try: 68 | import onnx 69 | 70 | print('\nStarting ONNX export with onnx %s...' % onnx.__version__) 71 | f = opt.weights.replace('.pt', '.onnx') # filename 72 | torch.onnx.export(model, img, f, verbose=False, opset_version=12, input_names=['images'], 73 | output_names=['classes', 'boxes'] if y is None else ['output']) 74 | 75 | # Checks 76 | onnx_model = onnx.load(f) # load onnx model 77 | onnx.checker.check_model(onnx_model) # check onnx model 78 | # print(onnx.helper.printable_graph(onnx_model.graph)) # print a human readable model 79 | print('ONNX export success, saved as %s' % f) 80 | except Exception as e: 81 | print('ONNX export failure: %s' % e) 82 | 83 | # CoreML export 84 | try: 85 | import coremltools as ct 86 | 87 | print('\nStarting CoreML export with coremltools %s...' % ct.__version__) 88 | # convert model from torchscript and apply pixel scaling as per detect.py 89 | model = ct.convert(ts, inputs=[ct.ImageType(name='image', shape=img.shape, scale=1 / 255.0, bias=[0, 0, 0])]) 90 | f = opt.weights.replace('.pt', '.mlmodel') # filename 91 | model.save(f) 92 | print('CoreML export success, saved as %s' % f) 93 | except Exception as e: 94 | print('CoreML export failure: %s' % e) 95 | 96 | # Finish 97 | print('\nExport complete (%.2fs). Visualize with https://github.com/lutzroeder/netron.' % (time.time() - t)) 98 | -------------------------------------------------------------------------------- /yolov5/models/hub/anchors.yaml: -------------------------------------------------------------------------------- 1 | # Default YOLOv5 anchors for COCO data 2 | 3 | 4 | # P5 ------------------------------------------------------------------------------------------------------------------- 5 | # P5-640: 6 | anchors_p5_640: 7 | - [ 10,13, 16,30, 33,23 ] # P3/8 8 | - [ 30,61, 62,45, 59,119 ] # P4/16 9 | - [ 116,90, 156,198, 373,326 ] # P5/32 10 | 11 | 12 | # P6 ------------------------------------------------------------------------------------------------------------------- 13 | # P6-640: thr=0.25: 0.9964 BPR, 5.54 anchors past thr, n=12, img_size=640, metric_all=0.281/0.716-mean/best, past_thr=0.469-mean: 9,11, 21,19, 17,41, 43,32, 39,70, 86,64, 65,131, 134,130, 120,265, 282,180, 247,354, 512,387 14 | anchors_p6_640: 15 | - [ 9,11, 21,19, 17,41 ] # P3/8 16 | - [ 43,32, 39,70, 86,64 ] # P4/16 17 | - [ 65,131, 134,130, 120,265 ] # P5/32 18 | - [ 282,180, 247,354, 512,387 ] # P6/64 19 | 20 | # P6-1280: thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1280, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 19,27, 44,40, 38,94, 96,68, 86,152, 180,137, 140,301, 303,264, 238,542, 436,615, 739,380, 925,792 21 | anchors_p6_1280: 22 | - [ 19,27, 44,40, 38,94 ] # P3/8 23 | - [ 96,68, 86,152, 180,137 ] # P4/16 24 | - [ 140,301, 303,264, 238,542 ] # P5/32 25 | - [ 436,615, 739,380, 925,792 ] # P6/64 26 | 27 | # P6-1920: thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1920, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 28,41, 67,59, 57,141, 144,103, 129,227, 270,205, 209,452, 455,396, 358,812, 653,922, 1109,570, 1387,1187 28 | anchors_p6_1920: 29 | - [ 28,41, 67,59, 57,141 ] # P3/8 30 | - [ 144,103, 129,227, 270,205 ] # P4/16 31 | - [ 209,452, 455,396, 358,812 ] # P5/32 32 | - [ 653,922, 1109,570, 1387,1187 ] # P6/64 33 | 34 | 35 | # P7 ------------------------------------------------------------------------------------------------------------------- 36 | # P7-640: thr=0.25: 0.9962 BPR, 6.76 anchors past thr, n=15, img_size=640, metric_all=0.275/0.733-mean/best, past_thr=0.466-mean: 11,11, 13,30, 29,20, 30,46, 61,38, 39,92, 78,80, 146,66, 79,163, 149,150, 321,143, 157,303, 257,402, 359,290, 524,372 37 | anchors_p7_640: 38 | - [ 11,11, 13,30, 29,20 ] # P3/8 39 | - [ 30,46, 61,38, 39,92 ] # P4/16 40 | - [ 78,80, 146,66, 79,163 ] # P5/32 41 | - [ 149,150, 321,143, 157,303 ] # P6/64 42 | - [ 257,402, 359,290, 524,372 ] # P7/128 43 | 44 | # P7-1280: thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1280, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 19,22, 54,36, 32,77, 70,83, 138,71, 75,173, 165,159, 148,334, 375,151, 334,317, 251,626, 499,474, 750,326, 534,814, 1079,818 45 | anchors_p7_1280: 46 | - [ 19,22, 54,36, 32,77 ] # P3/8 47 | - [ 70,83, 138,71, 75,173 ] # P4/16 48 | - [ 165,159, 148,334, 375,151 ] # P5/32 49 | - [ 334,317, 251,626, 499,474 ] # P6/64 50 | - [ 750,326, 534,814, 1079,818 ] # P7/128 51 | 52 | # P7-1920: thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1920, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 29,34, 81,55, 47,115, 105,124, 207,107, 113,259, 247,238, 222,500, 563,227, 501,476, 376,939, 749,711, 1126,489, 801,1222, 1618,1227 53 | anchors_p7_1920: 54 | - [ 29,34, 81,55, 47,115 ] # P3/8 55 | - [ 105,124, 207,107, 113,259 ] # P4/16 56 | - [ 247,238, 222,500, 563,227 ] # P5/32 57 | - [ 501,476, 376,939, 749,711 ] # P6/64 58 | - [ 1126,489, 801,1222, 1618,1227 ] # P7/128 59 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov3-spp.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # darknet53 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Conv, [32, 3, 1]], # 0 16 | [-1, 1, Conv, [64, 3, 2]], # 1-P1/2 17 | [-1, 1, Bottleneck, [64]], 18 | [-1, 1, Conv, [128, 3, 2]], # 3-P2/4 19 | [-1, 2, Bottleneck, [128]], 20 | [-1, 1, Conv, [256, 3, 2]], # 5-P3/8 21 | [-1, 8, Bottleneck, [256]], 22 | [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 23 | [-1, 8, Bottleneck, [512]], 24 | [-1, 1, Conv, [1024, 3, 2]], # 9-P5/32 25 | [-1, 4, Bottleneck, [1024]], # 10 26 | ] 27 | 28 | # YOLOv3-SPP head 29 | head: 30 | [[-1, 1, Bottleneck, [1024, False]], 31 | [-1, 1, SPP, [512, [5, 9, 13]]], 32 | [-1, 1, Conv, [1024, 3, 1]], 33 | [-1, 1, Conv, [512, 1, 1]], 34 | [-1, 1, Conv, [1024, 3, 1]], # 15 (P5/32-large) 35 | 36 | [-2, 1, Conv, [256, 1, 1]], 37 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 38 | [[-1, 8], 1, Concat, [1]], # cat backbone P4 39 | [-1, 1, Bottleneck, [512, False]], 40 | [-1, 1, Bottleneck, [512, False]], 41 | [-1, 1, Conv, [256, 1, 1]], 42 | [-1, 1, Conv, [512, 3, 1]], # 22 (P4/16-medium) 43 | 44 | [-2, 1, Conv, [128, 1, 1]], 45 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 46 | [[-1, 6], 1, Concat, [1]], # cat backbone P3 47 | [-1, 1, Bottleneck, [256, False]], 48 | [-1, 2, Bottleneck, [256, False]], # 27 (P3/8-small) 49 | 50 | [[27, 22, 15], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 51 | ] 52 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov3-tiny.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,14, 23,27, 37,58] # P4/16 9 | - [81,82, 135,169, 344,319] # P5/32 10 | 11 | # YOLOv3-tiny backbone 12 | backbone: 13 | # [from, number, module, args] 14 | [[-1, 1, Conv, [16, 3, 1]], # 0 15 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 1-P1/2 16 | [-1, 1, Conv, [32, 3, 1]], 17 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 3-P2/4 18 | [-1, 1, Conv, [64, 3, 1]], 19 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 5-P3/8 20 | [-1, 1, Conv, [128, 3, 1]], 21 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 7-P4/16 22 | [-1, 1, Conv, [256, 3, 1]], 23 | [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 9-P5/32 24 | [-1, 1, Conv, [512, 3, 1]], 25 | [-1, 1, nn.ZeroPad2d, [[0, 1, 0, 1]]], # 11 26 | [-1, 1, nn.MaxPool2d, [2, 1, 0]], # 12 27 | ] 28 | 29 | # YOLOv3-tiny head 30 | head: 31 | [[-1, 1, Conv, [1024, 3, 1]], 32 | [-1, 1, Conv, [256, 1, 1]], 33 | [-1, 1, Conv, [512, 3, 1]], # 15 (P5/32-large) 34 | 35 | [-2, 1, Conv, [128, 1, 1]], 36 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 37 | [[-1, 8], 1, Concat, [1]], # cat backbone P4 38 | [-1, 1, Conv, [256, 3, 1]], # 19 (P4/16-medium) 39 | 40 | [[19, 15], 1, Detect, [nc, anchors]], # Detect(P4, P5) 41 | ] 42 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov3.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # darknet53 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Conv, [32, 3, 1]], # 0 16 | [-1, 1, Conv, [64, 3, 2]], # 1-P1/2 17 | [-1, 1, Bottleneck, [64]], 18 | [-1, 1, Conv, [128, 3, 2]], # 3-P2/4 19 | [-1, 2, Bottleneck, [128]], 20 | [-1, 1, Conv, [256, 3, 2]], # 5-P3/8 21 | [-1, 8, Bottleneck, [256]], 22 | [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 23 | [-1, 8, Bottleneck, [512]], 24 | [-1, 1, Conv, [1024, 3, 2]], # 9-P5/32 25 | [-1, 4, Bottleneck, [1024]], # 10 26 | ] 27 | 28 | # YOLOv3 head 29 | head: 30 | [[-1, 1, Bottleneck, [1024, False]], 31 | [-1, 1, Conv, [512, [1, 1]]], 32 | [-1, 1, Conv, [1024, 3, 1]], 33 | [-1, 1, Conv, [512, 1, 1]], 34 | [-1, 1, Conv, [1024, 3, 1]], # 15 (P5/32-large) 35 | 36 | [-2, 1, Conv, [256, 1, 1]], 37 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 38 | [[-1, 8], 1, Concat, [1]], # cat backbone P4 39 | [-1, 1, Bottleneck, [512, False]], 40 | [-1, 1, Bottleneck, [512, False]], 41 | [-1, 1, Conv, [256, 1, 1]], 42 | [-1, 1, Conv, [512, 3, 1]], # 22 (P4/16-medium) 43 | 44 | [-2, 1, Conv, [128, 1, 1]], 45 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 46 | [[-1, 6], 1, Concat, [1]], # cat backbone P3 47 | [-1, 1, Bottleneck, [256, False]], 48 | [-1, 2, Bottleneck, [256, False]], # 27 (P3/8-small) 49 | 50 | [[27, 22, 15], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 51 | ] 52 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov5-fpn.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, Bottleneck, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, BottleneckCSP, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, BottleneckCSP, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 6, BottleneckCSP, [1024]], # 9 25 | ] 26 | 27 | # YOLOv5 FPN head 28 | head: 29 | [[-1, 3, BottleneckCSP, [1024, False]], # 10 (P5/32-large) 30 | 31 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 32 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 33 | [-1, 1, Conv, [512, 1, 1]], 34 | [-1, 3, BottleneckCSP, [512, False]], # 14 (P4/16-medium) 35 | 36 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 37 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 38 | [-1, 1, Conv, [256, 1, 1]], 39 | [-1, 3, BottleneckCSP, [256, False]], # 18 (P3/8-small) 40 | 41 | [[18, 14, 10], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 42 | ] 43 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov5-p2.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 3 8 | 9 | # YOLOv5 backbone 10 | backbone: 11 | # [from, number, module, args] 12 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 13 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 14 | [ -1, 3, C3, [ 128 ] ], 15 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 16 | [ -1, 9, C3, [ 256 ] ], 17 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 18 | [ -1, 9, C3, [ 512 ] ], 19 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 7-P5/32 20 | [ -1, 1, SPP, [ 1024, [ 5, 9, 13 ] ] ], 21 | [ -1, 3, C3, [ 1024, False ] ], # 9 22 | ] 23 | 24 | # YOLOv5 head 25 | head: 26 | [ [ -1, 1, Conv, [ 512, 1, 1 ] ], 27 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 28 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 29 | [ -1, 3, C3, [ 512, False ] ], # 13 30 | 31 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 32 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 33 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 34 | [ -1, 3, C3, [ 256, False ] ], # 17 (P3/8-small) 35 | 36 | [ -1, 1, Conv, [ 128, 1, 1 ] ], 37 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 38 | [ [ -1, 2 ], 1, Concat, [ 1 ] ], # cat backbone P2 39 | [ -1, 1, C3, [ 128, False ] ], # 21 (P2/4-xsmall) 40 | 41 | [ -1, 1, Conv, [ 128, 3, 2 ] ], 42 | [ [ -1, 18 ], 1, Concat, [ 1 ] ], # cat head P3 43 | [ -1, 3, C3, [ 256, False ] ], # 24 (P3/8-small) 44 | 45 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 46 | [ [ -1, 14 ], 1, Concat, [ 1 ] ], # cat head P4 47 | [ -1, 3, C3, [ 512, False ] ], # 27 (P4/16-medium) 48 | 49 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 50 | [ [ -1, 10 ], 1, Concat, [ 1 ] ], # cat head P5 51 | [ -1, 3, C3, [ 1024, False ] ], # 30 (P5/32-large) 52 | 53 | [ [ 24, 27, 30 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5) 54 | ] 55 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov5-p6.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 3 8 | 9 | # YOLOv5 backbone 10 | backbone: 11 | # [from, number, module, args] 12 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 13 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 14 | [ -1, 3, C3, [ 128 ] ], 15 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 16 | [ -1, 9, C3, [ 256 ] ], 17 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 18 | [ -1, 9, C3, [ 512 ] ], 19 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 20 | [ -1, 3, C3, [ 768 ] ], 21 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 22 | [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ], 23 | [ -1, 3, C3, [ 1024, False ] ], # 11 24 | ] 25 | 26 | # YOLOv5 head 27 | head: 28 | [ [ -1, 1, Conv, [ 768, 1, 1 ] ], 29 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 30 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 31 | [ -1, 3, C3, [ 768, False ] ], # 15 32 | 33 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 34 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 35 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 36 | [ -1, 3, C3, [ 512, False ] ], # 19 37 | 38 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 39 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 40 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 41 | [ -1, 3, C3, [ 256, False ] ], # 23 (P3/8-small) 42 | 43 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 44 | [ [ -1, 20 ], 1, Concat, [ 1 ] ], # cat head P4 45 | [ -1, 3, C3, [ 512, False ] ], # 26 (P4/16-medium) 46 | 47 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 48 | [ [ -1, 16 ], 1, Concat, [ 1 ] ], # cat head P5 49 | [ -1, 3, C3, [ 768, False ] ], # 29 (P5/32-large) 50 | 51 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 52 | [ [ -1, 12 ], 1, Concat, [ 1 ] ], # cat head P6 53 | [ -1, 3, C3, [ 1024, False ] ], # 32 (P5/64-xlarge) 54 | 55 | [ [ 23, 26, 29, 32 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6) 56 | ] 57 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov5-p7.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 3 8 | 9 | # YOLOv5 backbone 10 | backbone: 11 | # [from, number, module, args] 12 | [ [ -1, 1, Focus, [ 64, 3 ] ], # 0-P1/2 13 | [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 14 | [ -1, 3, C3, [ 128 ] ], 15 | [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 16 | [ -1, 9, C3, [ 256 ] ], 17 | [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 18 | [ -1, 9, C3, [ 512 ] ], 19 | [ -1, 1, Conv, [ 768, 3, 2 ] ], # 7-P5/32 20 | [ -1, 3, C3, [ 768 ] ], 21 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 9-P6/64 22 | [ -1, 3, C3, [ 1024 ] ], 23 | [ -1, 1, Conv, [ 1280, 3, 2 ] ], # 11-P7/128 24 | [ -1, 1, SPP, [ 1280, [ 3, 5 ] ] ], 25 | [ -1, 3, C3, [ 1280, False ] ], # 13 26 | ] 27 | 28 | # YOLOv5 head 29 | head: 30 | [ [ -1, 1, Conv, [ 1024, 1, 1 ] ], 31 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 32 | [ [ -1, 10 ], 1, Concat, [ 1 ] ], # cat backbone P6 33 | [ -1, 3, C3, [ 1024, False ] ], # 17 34 | 35 | [ -1, 1, Conv, [ 768, 1, 1 ] ], 36 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 37 | [ [ -1, 8 ], 1, Concat, [ 1 ] ], # cat backbone P5 38 | [ -1, 3, C3, [ 768, False ] ], # 21 39 | 40 | [ -1, 1, Conv, [ 512, 1, 1 ] ], 41 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 42 | [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 43 | [ -1, 3, C3, [ 512, False ] ], # 25 44 | 45 | [ -1, 1, Conv, [ 256, 1, 1 ] ], 46 | [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], 47 | [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 48 | [ -1, 3, C3, [ 256, False ] ], # 29 (P3/8-small) 49 | 50 | [ -1, 1, Conv, [ 256, 3, 2 ] ], 51 | [ [ -1, 26 ], 1, Concat, [ 1 ] ], # cat head P4 52 | [ -1, 3, C3, [ 512, False ] ], # 32 (P4/16-medium) 53 | 54 | [ -1, 1, Conv, [ 512, 3, 2 ] ], 55 | [ [ -1, 22 ], 1, Concat, [ 1 ] ], # cat head P5 56 | [ -1, 3, C3, [ 768, False ] ], # 35 (P5/32-large) 57 | 58 | [ -1, 1, Conv, [ 768, 3, 2 ] ], 59 | [ [ -1, 18 ], 1, Concat, [ 1 ] ], # cat head P6 60 | [ -1, 3, C3, [ 1024, False ] ], # 38 (P6/64-xlarge) 61 | 62 | [ -1, 1, Conv, [ 1024, 3, 2 ] ], 63 | [ [ -1, 14 ], 1, Concat, [ 1 ] ], # cat head P7 64 | [ -1, 3, C3, [ 1280, False ] ], # 41 (P7/128-xxlarge) 65 | 66 | [ [ 29, 32, 35, 38, 41 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4, P5, P6, P7) 67 | ] 68 | -------------------------------------------------------------------------------- /yolov5/models/hub/yolov5-panet.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, BottleneckCSP, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, BottleneckCSP, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, BottleneckCSP, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, BottleneckCSP, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 PANet head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, BottleneckCSP, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, BottleneckCSP, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, BottleneckCSP, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, BottleneckCSP, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /yolov5/models/yolo.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | import logging 3 | import sys 4 | from copy import deepcopy 5 | from pathlib import Path 6 | 7 | sys.path.append('./') # to run '$ python *.py' files in subdirectories 8 | logger = logging.getLogger(__name__) 9 | 10 | from models.common import * 11 | from models.experimental import MixConv2d, CrossConv 12 | from utils.autoanchor import check_anchor_order 13 | from utils.general import make_divisible, check_file, set_logging 14 | from utils.torch_utils import time_synchronized, fuse_conv_and_bn, model_info, scale_img, initialize_weights, \ 15 | select_device, copy_attr 16 | 17 | try: 18 | import thop # for FLOPS computation 19 | except ImportError: 20 | thop = None 21 | 22 | 23 | class Detect(nn.Module): 24 | stride = None # strides computed during build 25 | export = False # onnx export 26 | 27 | def __init__(self, nc=80, anchors=(), ch=()): # detection layer 28 | super(Detect, self).__init__() 29 | self.nc = nc # number of classes 30 | self.no = nc + 5 # number of outputs per anchor 31 | self.nl = len(anchors) # number of detection layers 32 | self.na = len(anchors[0]) // 2 # number of anchors 33 | self.grid = [torch.zeros(1)] * self.nl # init grid 34 | a = torch.tensor(anchors).float().view(self.nl, -1, 2) 35 | self.register_buffer('anchors', a) # shape(nl,na,2) 36 | self.register_buffer('anchor_grid', a.clone().view(self.nl, 1, -1, 1, 1, 2)) # shape(nl,1,na,1,1,2) 37 | self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch) # output conv 38 | 39 | def forward(self, x): 40 | # x = x.copy() # for profiling 41 | z = [] # inference output 42 | self.training |= self.export 43 | for i in range(self.nl): 44 | x[i] = self.m[i](x[i]) # conv 45 | bs, _, ny, nx = x[i].shape # x(bs,255,20,20) to x(bs,3,20,20,85) 46 | x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous() 47 | 48 | if not self.training: # inference 49 | if self.grid[i].shape[2:4] != x[i].shape[2:4]: 50 | self.grid[i] = self._make_grid(nx, ny).to(x[i].device) 51 | 52 | y = x[i].sigmoid() 53 | y[..., 0:2] = (y[..., 0:2] * 2. - 0.5 + self.grid[i].to(x[i].device)) * self.stride[i] # xy 54 | y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] # wh 55 | z.append(y.view(bs, -1, self.no)) 56 | 57 | return x if self.training else (torch.cat(z, 1), x) 58 | 59 | @staticmethod 60 | def _make_grid(nx=20, ny=20): 61 | yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)]) 62 | return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float() 63 | 64 | 65 | class Model(nn.Module): 66 | def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None): # model, input channels, number of classes 67 | super(Model, self).__init__() 68 | if isinstance(cfg, dict): 69 | self.yaml = cfg # model dict 70 | else: # is *.yaml 71 | import yaml # for torch hub 72 | self.yaml_file = Path(cfg).name 73 | with open(cfg) as f: 74 | self.yaml = yaml.load(f, Loader=yaml.FullLoader) # model dict 75 | 76 | # Define model 77 | ch = self.yaml['ch'] = self.yaml.get('ch', ch) # input channels 78 | if nc and nc != self.yaml['nc']: 79 | logger.info('Overriding model.yaml nc=%g with nc=%g' % (self.yaml['nc'], nc)) 80 | self.yaml['nc'] = nc # override yaml value 81 | self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # model, savelist 82 | self.names = [str(i) for i in range(self.yaml['nc'])] # default names 83 | # print([x.shape for x in self.forward(torch.zeros(1, ch, 64, 64))]) 84 | 85 | # Build strides, anchors 86 | m = self.model[-1] # Detect() 87 | if isinstance(m, Detect): 88 | s = 256 # 2x min stride 89 | m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward 90 | m.anchors /= m.stride.view(-1, 1, 1) 91 | check_anchor_order(m) 92 | self.stride = m.stride 93 | self._initialize_biases() # only run once 94 | # print('Strides: %s' % m.stride.tolist()) 95 | 96 | # Init weights, biases 97 | initialize_weights(self) 98 | self.info() 99 | logger.info('') 100 | 101 | def forward(self, x, augment=False, profile=False): 102 | if augment: 103 | img_size = x.shape[-2:] # height, width 104 | s = [1, 0.83, 0.67] # scales 105 | f = [None, 3, None] # flips (2-ud, 3-lr) 106 | y = [] # outputs 107 | for si, fi in zip(s, f): 108 | xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max())) 109 | yi = self.forward_once(xi)[0] # forward 110 | # cv2.imwrite('img%g.jpg' % s, 255 * xi[0].numpy().transpose((1, 2, 0))[:, :, ::-1]) # save 111 | yi[..., :4] /= si # de-scale 112 | if fi == 2: 113 | yi[..., 1] = img_size[0] - yi[..., 1] # de-flip ud 114 | elif fi == 3: 115 | yi[..., 0] = img_size[1] - yi[..., 0] # de-flip lr 116 | y.append(yi) 117 | return torch.cat(y, 1), None # augmented inference, train 118 | else: 119 | return self.forward_once(x, profile) # single-scale inference, train 120 | 121 | def forward_once(self, x, profile=False): 122 | y, dt = [], [] # outputs 123 | for m in self.model: 124 | if m.f != -1: # if not from previous layer 125 | 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 126 | 127 | if profile: 128 | o = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 if thop else 0 # FLOPS 129 | t = time_synchronized() 130 | for _ in range(10): 131 | _ = m(x) 132 | dt.append((time_synchronized() - t) * 100) 133 | print('%10.1f%10.0f%10.1fms %-40s' % (o, m.np, dt[-1], m.type)) 134 | 135 | x = m(x) # run 136 | y.append(x if m.i in self.save else None) # save output 137 | 138 | if profile: 139 | print('%.1fms total' % sum(dt)) 140 | return x 141 | 142 | def _initialize_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency 143 | # https://arxiv.org/abs/1708.02002 section 3.3 144 | # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. 145 | m = self.model[-1] # Detect() module 146 | for mi, s in zip(m.m, m.stride): # from 147 | b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) 148 | b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) 149 | b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls 150 | mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) 151 | 152 | def _print_biases(self): 153 | m = self.model[-1] # Detect() module 154 | for mi in m.m: # from 155 | b = mi.bias.detach().view(m.na, -1).T # conv.bias(255) to (3,85) 156 | print(('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean())) 157 | 158 | # def _print_weights(self): 159 | # for m in self.model.modules(): 160 | # if type(m) is Bottleneck: 161 | # print('%10.3g' % (m.w.detach().sigmoid() * 2)) # shortcut weights 162 | 163 | def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers 164 | print('Fusing layers... ') 165 | for m in self.model.modules(): 166 | if type(m) is Conv and hasattr(m, 'bn'): 167 | m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv 168 | delattr(m, 'bn') # remove batchnorm 169 | m.forward = m.fuseforward # update forward 170 | self.info() 171 | return self 172 | 173 | def nms(self, mode=True): # add or remove NMS module 174 | present = type(self.model[-1]) is NMS # last layer is NMS 175 | if mode and not present: 176 | print('Adding NMS... ') 177 | m = NMS() # module 178 | m.f = -1 # from 179 | m.i = self.model[-1].i + 1 # index 180 | self.model.add_module(name='%s' % m.i, module=m) # add 181 | self.eval() 182 | elif not mode and present: 183 | print('Removing NMS... ') 184 | self.model = self.model[:-1] # remove 185 | return self 186 | 187 | def autoshape(self): # add autoShape module 188 | print('Adding autoShape... ') 189 | m = autoShape(self) # wrap model 190 | copy_attr(m, self, include=('yaml', 'nc', 'hyp', 'names', 'stride'), exclude=()) # copy attributes 191 | return m 192 | 193 | def info(self, verbose=False, img_size=640): # print model information 194 | model_info(self, verbose, img_size) 195 | 196 | 197 | def parse_model(d, ch): # model_dict, input_channels(3) 198 | logger.info('\n%3s%18s%3s%10s %-40s%-30s' % ('', 'from', 'n', 'params', 'module', 'arguments')) 199 | anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple'] 200 | na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors # number of anchors 201 | no = na * (nc + 5) # number of outputs = anchors * (classes + 5) 202 | 203 | layers, save, c2 = [], [], ch[-1] # layers, savelist, ch out 204 | for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']): # from, number, module, args 205 | m = eval(m) if isinstance(m, str) else m # eval strings 206 | for j, a in enumerate(args): 207 | try: 208 | args[j] = eval(a) if isinstance(a, str) else a # eval strings 209 | except: 210 | pass 211 | 212 | n = max(round(n * gd), 1) if n > 1 else n # depth gain 213 | if m in [Conv, Bottleneck, SPP, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3]: 214 | c1, c2 = ch[f], args[0] 215 | 216 | # Normal 217 | # if i > 0 and args[0] != no: # channel expansion factor 218 | # ex = 1.75 # exponential (default 2.0) 219 | # e = math.log(c2 / ch[1]) / math.log(2) 220 | # c2 = int(ch[1] * ex ** e) 221 | # if m != Focus: 222 | 223 | c2 = make_divisible(c2 * gw, 8) if c2 != no else c2 224 | 225 | # Experimental 226 | # if i > 0 and args[0] != no: # channel expansion factor 227 | # ex = 1 + gw # exponential (default 2.0) 228 | # ch1 = 32 # ch[1] 229 | # e = math.log(c2 / ch1) / math.log(2) # level 1-n 230 | # c2 = int(ch1 * ex ** e) 231 | # if m != Focus: 232 | # c2 = make_divisible(c2, 8) if c2 != no else c2 233 | 234 | args = [c1, c2, *args[1:]] 235 | if m in [BottleneckCSP, C3]: 236 | args.insert(2, n) 237 | n = 1 238 | elif m is nn.BatchNorm2d: 239 | args = [ch[f]] 240 | elif m is Concat: 241 | c2 = sum([ch[x if x < 0 else x + 1] for x in f]) 242 | elif m is Detect: 243 | args.append([ch[x + 1] for x in f]) 244 | if isinstance(args[1], int): # number of anchors 245 | args[1] = [list(range(args[1] * 2))] * len(f) 246 | elif m is Contract: 247 | c2 = ch[f if f < 0 else f + 1] * args[0] ** 2 248 | elif m is Expand: 249 | c2 = ch[f if f < 0 else f + 1] // args[0] ** 2 250 | else: 251 | c2 = ch[f if f < 0 else f + 1] 252 | 253 | m_ = nn.Sequential(*[m(*args) for _ in range(n)]) if n > 1 else m(*args) # module 254 | t = str(m)[8:-2].replace('__main__.', '') # module type 255 | np = sum([x.numel() for x in m_.parameters()]) # number params 256 | m_.i, m_.f, m_.type, m_.np = i, f, t, np # attach index, 'from' index, type, number params 257 | logger.info('%3s%18s%3s%10.0f %-40s%-30s' % (i, f, n, np, t, args)) # print 258 | save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1) # append to savelist 259 | layers.append(m_) 260 | ch.append(c2) 261 | return nn.Sequential(*layers), sorted(save) 262 | 263 | 264 | if __name__ == '__main__': 265 | parser = argparse.ArgumentParser() 266 | parser.add_argument('--cfg', type=str, default='yolov5s.yaml', help='model.yaml') 267 | parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') 268 | opt = parser.parse_args() 269 | opt.cfg = check_file(opt.cfg) # check file 270 | set_logging() 271 | device = select_device(opt.device) 272 | 273 | # Create model 274 | model = Model(opt.cfg).to(device) 275 | model.train() 276 | 277 | # Profile 278 | # img = torch.rand(8 if torch.cuda.is_available() else 1, 3, 640, 640).to(device) 279 | # y = model(img, profile=True) 280 | 281 | # Tensorboard 282 | # from torch.utils.tensorboard import SummaryWriter 283 | # tb_writer = SummaryWriter() 284 | # print("Run 'tensorboard --logdir=models/runs' to view tensorboard at http://localhost:6006/") 285 | # tb_writer.add_graph(model.model, img) # add model to tensorboard 286 | # tb_writer.add_image('test', img[0], dataformats='CWH') # add model to tensorboard 287 | -------------------------------------------------------------------------------- /yolov5/models/yolov5l.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.0 # model depth multiple 4 | width_multiple: 1.0 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /yolov5/models/yolov5m.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.67 # model depth multiple 4 | width_multiple: 0.75 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /yolov5/models/yolov5s.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 0.33 # model depth multiple 4 | width_multiple: 0.50 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | -------------------------------------------------------------------------------- /yolov5/models/yolov5x.yaml: -------------------------------------------------------------------------------- 1 | # parameters 2 | nc: 80 # number of classes 3 | depth_multiple: 1.33 # model depth multiple 4 | width_multiple: 1.25 # layer channel multiple 5 | 6 | # anchors 7 | anchors: 8 | - [10,13, 16,30, 33,23] # P3/8 9 | - [30,61, 62,45, 59,119] # P4/16 10 | - [116,90, 156,198, 373,326] # P5/32 11 | 12 | # YOLOv5 backbone 13 | backbone: 14 | # [from, number, module, args] 15 | [[-1, 1, Focus, [64, 3]], # 0-P1/2 16 | [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 17 | [-1, 3, C3, [128]], 18 | [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 19 | [-1, 9, C3, [256]], 20 | [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 21 | [-1, 9, C3, [512]], 22 | [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 23 | [-1, 1, SPP, [1024, [5, 9, 13]]], 24 | [-1, 3, C3, [1024, False]], # 9 25 | ] 26 | 27 | # YOLOv5 head 28 | head: 29 | [[-1, 1, Conv, [512, 1, 1]], 30 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 31 | [[-1, 6], 1, Concat, [1]], # cat backbone P4 32 | [-1, 3, C3, [512, False]], # 13 33 | 34 | [-1, 1, Conv, [256, 1, 1]], 35 | [-1, 1, nn.Upsample, [None, 2, 'nearest']], 36 | [[-1, 4], 1, Concat, [1]], # cat backbone P3 37 | [-1, 3, C3, [256, False]], # 17 (P3/8-small) 38 | 39 | [-1, 1, Conv, [256, 3, 2]], 40 | [[-1, 14], 1, Concat, [1]], # cat head P4 41 | [-1, 3, C3, [512, False]], # 20 (P4/16-medium) 42 | 43 | [-1, 1, Conv, [512, 3, 2]], 44 | [[-1, 10], 1, Concat, [1]], # cat head P5 45 | [-1, 3, C3, [1024, False]], # 23 (P5/32-large) 46 | 47 | [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) 48 | ] 49 | --------------------------------------------------------------------------------