# Yolo v4, v3 and v2 for Windows and Linux * Read the FAQ: https://www.ccoderun.ca/programming/darknet_faq/ * Join the Darknet/YOLO Discord: https://discord.gg/zSq8rtW * Recommended GitHub repo for Darknet/YOLO: https://github.com/hank-ai/darknetcv/ * Hank.ai and Darknet/YOLO: https://hank.ai/darknet-welcomes-hank-ai-as-official-sponsor-and-commercial-entity/ ## (neural networks for object detection) * Paper **YOLOv7**: https://arxiv.org/abs/2207.02696 * source code YOLOv7 - Pytorch (use to reproduce results): https://github.com/WongKinYiu/yolov7 ---- * Paper **YOLOv4**: https://arxiv.org/abs/2004.10934 * source code YOLOv4 - Darknet (use to reproduce results): https://github.com/AlexeyAB/darknet ---- * Paper **Scaled-YOLOv4 (CVPR 2021)**: https://openaccess.thecvf.com/content/CVPR2021/html/Wang_Scaled-YOLOv4_Scaling_Cross_Stage_Partial_Network_CVPR_2021_paper.html * source code Scaled-YOLOv4 - Pytorch (use to reproduce results): https://github.com/WongKinYiu/ScaledYOLOv4 ---- ### YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors * **Paper**: https://arxiv.org/abs/2207.02696 * **source code - Pytorch (use to reproduce results):** https://github.com/WongKinYiu/yolov7 YOLOv7 is more accurate and faster than YOLOv5 by **120%** FPS, than YOLOX by **180%** FPS, than Dual-Swin-T by **1200%** FPS, than ConvNext by **550%** FPS, than SWIN-L by **500%** FPS, than PPYOLOE-X by **150%** FPS. YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 160 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100, batch=1. * YOLOv7-e6 (55.9% AP, 56 FPS V100 b=1) by `+500%` FPS faster than SWIN-L C-M-RCNN (53.9% AP, 9.2 FPS A100 b=1) * YOLOv7-e6 (55.9% AP, 56 FPS V100 b=1) by `+550%` FPS faster than ConvNeXt-XL C-M-RCNN (55.2% AP, 8.6 FPS A100 b=1) * YOLOv7-w6 (54.6% AP, 84 FPS V100 b=1) by `+120%` FPS faster than YOLOv5-X6-r6.1 (55.0% AP, 38 FPS V100 b=1) * YOLOv7-w6 (54.6% AP, 84 FPS V100 b=1) by `+1200%` FPS faster than Dual-Swin-T C-M-RCNN (53.6% AP, 6.5 FPS V100 b=1) * YOLOv7x (52.9% AP, 114 FPS V100 b=1) by `+150%` FPS faster than PPYOLOE-X (51.9% AP, 45 FPS V100 b=1) * YOLOv7 (51.2% AP, 161 FPS V100 b=1) by `+180%` FPS faster than YOLOX-X (51.1% AP, 58 FPS V100 b=1) ---- ![more5](https://user-images.githubusercontent.com/4096485/179425274-f55a36d4-8450-4471-816b-8c105841effd.jpg) ---- ![image](https://user-images.githubusercontent.com/4096485/177675030-a929ee00-0eba-4d93-95c2-225231d0fd61.png) ---- More details in articles on medium: - [Scaled_YOLOv4](https://alexeyab84.medium.com/scaled-yolo-v4-is-the-best-neural-network-for-object-detection-on-ms-coco-dataset-39dfa22fa982?source=friends_link&sk=c8553bfed861b1a7932f739d26f487c8) - [YOLOv4](https://medium.com/@alexeyab84/yolov4-the-most-accurate-real-time-neural-network-on-ms-coco-dataset-73adfd3602fe?source=friends_link&sk=6039748846bbcf1d960c3061542591d7) Manual: https://github.com/AlexeyAB/darknet/wiki Discussion: - [Discord](https://discord.gg/zSq8rtW) About Darknet framework: http://pjreddie.com/darknet/ [![Darknet Continuous Integration](https://github.com/AlexeyAB/darknet/workflows/Darknet%20Continuous%20Integration/badge.svg)](https://github.com/AlexeyAB/darknet/actions?query=workflow%3A%22Darknet+Continuous+Integration%22) [![CircleCI](https://circleci.com/gh/AlexeyAB/darknet.svg?style=svg)](https://circleci.com/gh/AlexeyAB/darknet) [![Contributors](https://img.shields.io/github/contributors/AlexeyAB/Darknet.svg)](https://github.com/AlexeyAB/darknet/graphs/contributors) [![License: Unlicense](https://img.shields.io/badge/license-Unlicense-blue.svg)](https://github.com/AlexeyAB/darknet/blob/master/LICENSE) [![DOI](https://zenodo.org/badge/75388965.svg)](https://zenodo.org/badge/latestdoi/75388965) [![arxiv.org](http://img.shields.io/badge/cs.CV-arXiv%3A2004.10934-B31B1B.svg)](https://arxiv.org/abs/2004.10934) [![arxiv.org](http://img.shields.io/badge/cs.CV-arXiv%3A2011.08036-B31B1B.svg)](https://arxiv.org/abs/2011.08036) [![colab](https://user-images.githubusercontent.com/4096485/86174089-b2709f80-bb29-11ea-9faf-3d8dc668a1a5.png)](https://colab.research.google.com/drive/12QusaaRj_lUwCGDvQNfICpa7kA7_a2dE) [![colab](https://user-images.githubusercontent.com/4096485/86174097-b56b9000-bb29-11ea-9240-c17f6bacfc34.png)](https://colab.research.google.com/drive/1_GdoqCJWXsChrOiY8sZMr_zbr_fH-0Fg) - [YOLOv4 model zoo](https://github.com/AlexeyAB/darknet/wiki/YOLOv4-model-zoo) - [Requirements (and how to install dependencies)](#requirements-for-windows-linux-and-macos) - [Pre-trained models](#pre-trained-models) - [FAQ - frequently asked questions](https://github.com/AlexeyAB/darknet/wiki/FAQ---frequently-asked-questions) - [Explanations in issues](https://github.com/AlexeyAB/darknet/issues?q=is%3Aopen+is%3Aissue+label%3AExplanations) - [Yolo v4 in other frameworks (TensorRT, TensorFlow, PyTorch, OpenVINO, OpenCV-dnn, TVM,...)](#yolo-v4-in-other-frameworks) - [Datasets](#datasets) - [Yolo v4, v3 and v2 for Windows and Linux](#yolo-v4-v3-and-v2-for-windows-and-linux) - [(neural networks for object detection)](#neural-networks-for-object-detection) - [GeForce RTX 2080 Ti](#geforce-rtx-2080-ti) - [Youtube video of results](#youtube-video-of-results) - [How to evaluate AP of YOLOv4 on the MS COCO evaluation server](#how-to-evaluate-ap-of-yolov4-on-the-ms-coco-evaluation-server) - [How to evaluate FPS of YOLOv4 on GPU](#how-to-evaluate-fps-of-yolov4-on-gpu) - [Pre-trained models](#pre-trained-models) - [Requirements for Windows, Linux and macOS](#requirements-for-windows-linux-and-macos) - [Yolo v4 in other frameworks](#yolo-v4-in-other-frameworks) - [Datasets](#datasets) - [Improvements in this repository](#improvements-in-this-repository) - [How to use on the command line](#how-to-use-on-the-command-line) - [For using network video-camera mjpeg-stream with any Android smartphone](#for-using-network-video-camera-mjpeg-stream-with-any-android-smartphone) - [How to compile on Linux/macOS (using `CMake`)](#how-to-compile-on-linuxmacos-using-cmake) - [Using also PowerShell](#using-also-powershell) - [How to compile on Linux (using `make`)](#how-to-compile-on-linux-using-make) - [How to compile on Windows (using `CMake`)](#how-to-compile-on-windows-using-cmake) - [How to compile on Windows (using `vcpkg`)](#how-to-compile-on-windows-using-vcpkg) - [How to train with multi-GPU](#how-to-train-with-multi-gpu) - [How to train (to detect your custom objects)](#how-to-train-to-detect-your-custom-objects) - [How to train tiny-yolo (to detect your custom objects)](#how-to-train-tiny-yolo-to-detect-your-custom-objects) - [When should I stop training](#when-should-i-stop-training) - [Custom object detection](#custom-object-detection) - [How to improve object detection](#how-to-improve-object-detection) - [How to mark bounded boxes of objects and create annotation files](#how-to-mark-bounded-boxes-of-objects-and-create-annotation-files) - [How to use Yolo as DLL and SO libraries](#how-to-use-yolo-as-dll-and-so-libraries) - [Citation](#citation) ![Darknet Logo](http://pjreddie.com/media/files/darknet-black-small.png) ![scaled_yolov4](https://user-images.githubusercontent.com/4096485/112776361-281d8380-9048-11eb-8083-8728b12dcd55.png) AP50:95 - FPS (Tesla V100) Paper: https://arxiv.org/abs/2011.08036 ---- ![modern_gpus](https://user-images.githubusercontent.com/4096485/82835867-f1c62380-9ecd-11ea-9134-1598ed2abc4b.png) AP50:95 / AP50 - FPS (Tesla V100) Paper: https://arxiv.org/abs/2004.10934 tkDNN-TensorRT accelerates YOLOv4 **~2x** times for batch=1 and **3x-4x** times for batch=4. - tkDNN: https://github.com/ceccocats/tkDNN - OpenCV: https://gist.github.com/YashasSamaga/48bdb167303e10f4d07b754888ddbdcf ### GeForce RTX 2080 Ti | Network Size | Darknet, FPS (avg) | tkDNN TensorRT FP32, FPS | tkDNN TensorRT FP16, FPS | OpenCV FP16, FPS | tkDNN TensorRT FP16 batch=4, FPS | OpenCV FP16 batch=4, FPS | tkDNN Speedup | |:--------------------------:|:------------------:|-------------------------:|-------------------------:|-----------------:|---------------------------------:|-------------------------:|--------------:| |320 | 100 | 116 | **202** | 183 | 423 | **430** | **4.3x** | |416 | 82 | 103 | **162** | 159 | 284 | **294** | **3.6x** | |512 | 69 | 91 | 134 | **138** | 206 | **216** | **3.1x** | |608 | 53 | 62 | 103 | **115** | 150 | **150** | **2.8x** | |Tiny 416 | 443 | 609 | **790** | 773 | **1774** | 1353 | **3.5x** | |Tiny 416 CPU Core i7 7700HQ | 3.4 | - | - | 42 | - | 39 | **12x** | - Yolo v4 Full comparison: [map_fps](https://user-images.githubusercontent.com/4096485/80283279-0e303e00-871f-11ea-814c-870967d77fd1.png) - Yolo v4 tiny comparison: [tiny_fps](https://user-images.githubusercontent.com/4096485/85734112-6e366700-b705-11ea-95d1-fcba0de76d72.png) - CSPNet: [paper](https://arxiv.org/abs/1911.11929) and [map_fps](https://user-images.githubusercontent.com/4096485/71702416-6645dc00-2de0-11ea-8d65-de7d4b604021.png) comparison: https://github.com/WongKinYiu/CrossStagePartialNetworks - Yolo v3 on MS COCO: [Speed / Accuracy (mAP@0.5) chart](https://user-images.githubusercontent.com/4096485/52151356-e5d4a380-2683-11e9-9d7d-ac7bc192c477.jpg) - Yolo v3 on MS COCO (Yolo v3 vs RetinaNet) - Figure 3: https://arxiv.org/pdf/1804.02767v1.pdf - Yolo v2 on Pascal VOC 2007: https://hsto.org/files/a24/21e/068/a2421e0689fb43f08584de9d44c2215f.jpg - Yolo v2 on Pascal VOC 2012 (comp4): https://hsto.org/files/3a6/fdf/b53/3a6fdfb533f34cee9b52bdd9bb0b19d9.jpg #### Youtube video of results | [![Yolo v4](https://user-images.githubusercontent.com/4096485/101360000-1a33cf00-38ae-11eb-9e5e-b29c5fb0afbe.png)](https://youtu.be/1_SiUOYUoOI "Yolo v4") | [![Scaled Yolo v4](https://user-images.githubusercontent.com/4096485/101359389-43a02b00-38ad-11eb-866c-f813e96bf61a.png)](https://youtu.be/YDFf-TqJOFE "Scaled Yolo v4") | |---|---| Others: https://www.youtube.com/user/pjreddie/videos #### How to evaluate AP of YOLOv4 on the MS COCO evaluation server 1. Download and unzip test-dev2017 dataset from MS COCO server: http://images.cocodataset.org/zips/test2017.zip 2. Download list of images for Detection tasks and replace the paths with yours: https://raw.githubusercontent.com/AlexeyAB/darknet/master/scripts/testdev2017.txt 3. Download `yolov4.weights` file 245 MB: [yolov4.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights) (Google-drive mirror [yolov4.weights](https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT) ) 4. Content of the file `cfg/coco.data` should be ```ini classes= 80 train = /trainvalno5k.txt valid = /testdev2017.txt names = data/coco.names backup = backup eval=coco ``` 5. Create `/results/` folder near with `./darknet` executable file 6. Run validation: `./darknet detector valid cfg/coco.data cfg/yolov4.cfg yolov4.weights` 7. Rename the file `/results/coco_results.json` to `detections_test-dev2017_yolov4_results.json` and compress it to `detections_test-dev2017_yolov4_results.zip` 8. Submit file `detections_test-dev2017_yolov4_results.zip` to the MS COCO evaluation server for the `test-dev2019 (bbox)` #### How to evaluate FPS of YOLOv4 on GPU 1. Compile Darknet with `GPU=1 CUDNN=1 CUDNN_HALF=1 OPENCV=1` in the `Makefile` 2. Download `yolov4.weights` file 245 MB: [yolov4.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights) (Google-drive mirror [yolov4.weights](https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT) ) 3. Get any .avi/.mp4 video file (preferably not more than 1920x1080 to avoid bottlenecks in CPU performance) 4. Run one of two commands and look at the AVG FPS: - include video_capturing + NMS + drawing_bboxes: `./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights test.mp4 -dont_show -ext_output` - exclude video_capturing + NMS + drawing_bboxes: `./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights test.mp4 -benchmark` #### Pre-trained models There are weights-file for different cfg-files (trained for MS COCO dataset): FPS on RTX 2070 (R) and Tesla V100 (V): - [yolov4-p6.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-p6.cfg) - 1280x1280 - **72.1% mAP@0.5 (54.0% AP@0.5:0.95) - 32(V) FPS** - xxx BFlops (xxx FMA) - 487 MB: [yolov4-p6.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-p6.weights) - pre-trained weights for training: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-p6.conv.289 - [yolov4-p5.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-p5.cfg) - 896x896 - **70.0% mAP@0.5 (51.6% AP@0.5:0.95) - 43(V) FPS** - xxx BFlops (xxx FMA) - 271 MB: [yolov4-p5.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-p5.weights) - pre-trained weights for training: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-p5.conv.232 - [yolov4-csp-x-swish.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-csp-x-swish.cfg) - 640x640 - **69.9% mAP@0.5 (51.5% AP@0.5:0.95) - 23(R) FPS / 50(V) FPS** - 221 BFlops (110 FMA) - 381 MB: [yolov4-csp-x-swish.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp-x-swish.weights) - pre-trained weights for training: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp-x-swish.conv.192 - [yolov4-csp-swish.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-csp-swish.cfg) - 640x640 - **68.7% mAP@0.5 (50.0% AP@0.5:0.95) - 70(V) FPS** - 120 (60 FMA) - 202 MB: [yolov4-csp-swish.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp-swish.weights) - pre-trained weights for training: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp-swish.conv.164 - [yolov4x-mish.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4x-mish.cfg) - 640x640 - **68.5% mAP@0.5 (50.1% AP@0.5:0.95) - 23(R) FPS / 50(V) FPS** - 221 BFlops (110 FMA) - 381 MB: [yolov4x-mish.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4x-mish.weights) - pre-trained weights for training: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4x-mish.conv.166 - [yolov4-csp.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-csp.cfg) - 202 MB: [yolov4-csp.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp.weights) paper [Scaled Yolo v4](https://arxiv.org/abs/2011.08036) just change `width=` and `height=` parameters in `yolov4-csp.cfg` file and use the same `yolov4-csp.weights` file for all cases: - `width=640 height=640` in cfg: **67.4% mAP@0.5 (48.7% AP@0.5:0.95) - 70(V) FPS** - 120 (60 FMA) BFlops - `width=512 height=512` in cfg: **64.8% mAP@0.5 (46.2% AP@0.5:0.95) - 93(V) FPS** - 77 (39 FMA) BFlops - pre-trained weights for training: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-csp.conv.142 - [yolov4.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4.cfg) - 245 MB: [yolov4.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights) (Google-drive mirror [yolov4.weights](https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT) ) paper [Yolo v4](https://arxiv.org/abs/2004.10934) just change `width=` and `height=` parameters in `yolov4.cfg` file and use the same `yolov4.weights` file for all cases: - `width=608 height=608` in cfg: **65.7% mAP@0.5 (43.5% AP@0.5:0.95) - 34(R) FPS / 62(V) FPS** - 128.5 BFlops - `width=512 height=512` in cfg: **64.9% mAP@0.5 (43.0% AP@0.5:0.95) - 45(R) FPS / 83(V) FPS** - 91.1 BFlops - `width=416 height=416` in cfg: **62.8% mAP@0.5 (41.2% AP@0.5:0.95) - 55(R) FPS / 96(V) FPS** - 60.1 BFlops - `width=320 height=320` in cfg: **60% mAP@0.5 ( 38% AP@0.5:0.95) - 63(R) FPS / 123(V) FPS** - 35.5 BFlops - [yolov4-tiny.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-tiny.cfg) - **40.2% mAP@0.5 - 371(1080Ti) FPS / 330(RTX2070) FPS** - 6.9 BFlops - 23.1 MB: [yolov4-tiny.weights](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.weights) - [enet-coco.cfg (EfficientNetB0-Yolov3)](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/enet-coco.cfg) - **45.5% mAP@0.5 - 55(R) FPS** - 3.7 BFlops - 18.3 MB: [enetb0-coco_final.weights](https://drive.google.com/file/d/1FlHeQjWEQVJt0ay1PVsiuuMzmtNyv36m/view) - [yolov3-openimages.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3-openimages.cfg) - 247 MB - 18(R) FPS - OpenImages dataset: [yolov3-openimages.weights](https://pjreddie.com/media/files/yolov3-openimages.weights)
CLICK ME - Yolo v3 models - [csresnext50-panet-spp-original-optimal.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/csresnext50-panet-spp-original-optimal.cfg) - **65.4% mAP@0.5 (43.2% AP@0.5:0.95) - 32(R) FPS** - 100.5 BFlops - 217 MB: [csresnext50-panet-spp-original-optimal_final.weights](https://drive.google.com/open?id=1_NnfVgj0EDtb_WLNoXV8Mo7WKgwdYZCc) - [yolov3-spp.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3-spp.cfg) - **60.6% mAP@0.5 - 38(R) FPS** - 141.5 BFlops - 240 MB: [yolov3-spp.weights](https://pjreddie.com/media/files/yolov3-spp.weights) - [csresnext50-panet-spp.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/csresnext50-panet-spp.cfg) - **60.0% mAP@0.5 - 44 FPS** - 71.3 BFlops - 217 MB: [csresnext50-panet-spp_final.weights](https://drive.google.com/file/d/1aNXdM8qVy11nqTcd2oaVB3mf7ckr258-/view?usp=sharing) - [yolov3.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3.cfg) - **55.3% mAP@0.5 - 66(R) FPS** - 65.9 BFlops - 236 MB: [yolov3.weights](https://pjreddie.com/media/files/yolov3.weights) - [yolov3-tiny.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3-tiny.cfg) - **33.1% mAP@0.5 - 345(R) FPS** - 5.6 BFlops - 33.7 MB: [yolov3-tiny.weights](https://pjreddie.com/media/files/yolov3-tiny.weights) - [yolov3-tiny-prn.cfg](https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3-tiny-prn.cfg) - **33.1% mAP@0.5 - 370(R) FPS** - 3.5 BFlops - 18.8 MB: [yolov3-tiny-prn.weights](https://drive.google.com/file/d/18yYZWyKbo4XSDVyztmsEcF9B_6bxrhUY/view?usp=sharing)
CLICK ME - Yolo v2 models - `yolov2.cfg` (194 MB COCO Yolo v2) - requires 4 GB GPU-RAM: https://pjreddie.com/media/files/yolov2.weights - `yolo-voc.cfg` (194 MB VOC Yolo v2) - requires 4 GB GPU-RAM: http://pjreddie.com/media/files/yolo-voc.weights - `yolov2-tiny.cfg` (43 MB COCO Yolo v2) - requires 1 GB GPU-RAM: https://pjreddie.com/media/files/yolov2-tiny.weights - `yolov2-tiny-voc.cfg` (60 MB VOC Yolo v2) - requires 1 GB GPU-RAM: http://pjreddie.com/media/files/yolov2-tiny-voc.weights - `yolo9000.cfg` (186 MB Yolo9000-model) - requires 4 GB GPU-RAM: http://pjreddie.com/media/files/yolo9000.weights
Put it near compiled: darknet.exe You can get cfg-files by path: `darknet/cfg/` ### Requirements for Windows, Linux and macOS - **CMake >= 3.18**: https://cmake.org/download/ - **Powershell** (already installed on windows): https://docs.microsoft.com/en-us/powershell/scripting/install/installing-powershell - **CUDA >= 10.2**: https://developer.nvidia.com/cuda-toolkit-archive (on Linux do [Post-installation Actions](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#post-installation-actions)) - **OpenCV >= 2.4**: use your preferred package manager (brew, apt), build from source using [vcpkg](https://github.com/Microsoft/vcpkg) or download from [OpenCV official site](https://opencv.org/releases.html) (on Windows set system variable `OpenCV_DIR` = `C:\opencv\build` - where are the `include` and `x64` folders [image](https://user-images.githubusercontent.com/4096485/53249516-5130f480-36c9-11e9-8238-a6e82e48c6f2.png)) - **cuDNN >= 8.0.2** https://developer.nvidia.com/rdp/cudnn-archive (on **Linux** follow steps described here https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#installlinux-tar , on **Windows** follow steps described here https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#installwindows) - **GPU with CC >= 3.0**: https://en.wikipedia.org/wiki/CUDA#GPUs_supported ### Yolo v4 in other frameworks - **Pytorch - Scaled-YOLOv4:** https://github.com/WongKinYiu/ScaledYOLOv4 - **TensorFlow:** `pip install yolov4` YOLOv4 on TensorFlow 2.0 / TFlite / Android: https://github.com/hunglc007/tensorflow-yolov4-tflite Official TF models: https://github.com/tensorflow/models/tree/master/official/vision/beta/projects/yolo For YOLOv4 - convert `yolov4.weights`/`cfg` files to `yolov4.pb` by using [TNTWEN](https://github.com/TNTWEN/OpenVINO-YOLOV4) project, and to `yolov4.tflite` [TensorFlow-lite](https://www.tensorflow.org/lite/guide/get_started#2_convert_the_model_format) - **OpenCV** the fastest implementation of YOLOv4 for CPU (x86/ARM-Android), OpenCV can be compiled with [OpenVINO-backend](https://github.com/opencv/opencv/wiki/Intel's-Deep-Learning-Inference-Engine-backend) for running on (Myriad X / USB Neural Compute Stick / Arria FPGA), use `yolov4.weights`/`cfg` with: [C++ example](https://github.com/opencv/opencv/blob/8c25a8eb7b10fb50cda323ee6bec68aa1a9ce43c/samples/dnn/object_detection.cpp#L192-L221) or [Python example](https://github.com/opencv/opencv/blob/8c25a8eb7b10fb50cda323ee6bec68aa1a9ce43c/samples/dnn/object_detection.py#L129-L150) - **Intel OpenVINO 2021.2:** supports YOLOv4 (NPU Myriad X / USB Neural Compute Stick / Arria FPGA): https://devmesh.intel.com/projects/openvino-yolov4-49c756 read this [manual](https://github.com/TNTWEN/OpenVINO-YOLOV4) (old [manual](https://software.intel.com/en-us/articles/OpenVINO-Using-TensorFlow#converting-a-darknet-yolo-model) ) (for [Scaled-YOLOv4](https://github.com/WongKinYiu/ScaledYOLOv4/tree/yolov4-large) models use https://github.com/Chen-MingChang/pytorch_YOLO_OpenVINO_demo ) - **PyTorch > ONNX**: - [WongKinYiu/PyTorch_YOLOv4](https://github.com/WongKinYiu/PyTorch_YOLOv4) - [maudzung/3D-YOLOv4](https://github.com/maudzung/Complex-YOLOv4-Pytorch) - [Tianxiaomo/pytorch-YOLOv4](https://github.com/Tianxiaomo/pytorch-YOLOv4) - [YOLOv5](https://github.com/ultralytics/yolov5) - **ONNX** on Jetson for YOLOv4: https://developer.nvidia.com/blog/announcing-onnx-runtime-for-jetson/ and https://github.com/ttanzhiqiang/onnx_tensorrt_project - **nVidia Transfer Learning Toolkit (TLT>=3.0)** Training and Detection https://docs.nvidia.com/metropolis/TLT/tlt-user-guide/text/object_detection/yolo_v4.html - **TensorRT+tkDNN**: https://github.com/ceccocats/tkDNN#fps-results - **Deepstream 5.0 / TensorRT for YOLOv4** https://github.com/NVIDIA-AI-IOT/yolov4_deepstream or https://github.com/marcoslucianops/DeepStream-Yolo read [Yolo is natively supported in DeepStream 4.0](https://news.developer.nvidia.com/deepstream-sdk-4-now-available/) and [PDF](https://docs.nvidia.com/metropolis/deepstream/Custom_YOLO_Model_in_the_DeepStream_YOLO_App.pdf). Additionally [jkjung-avt/tensorrt_demos](https://github.com/jkjung-avt/tensorrt_demos) or [wang-xinyu/tensorrtx](https://github.com/wang-xinyu/tensorrtx) - **Triton Inference Server / TensorRT** https://github.com/isarsoft/yolov4-triton-tensorrt - **DirectML** https://github.com/microsoft/DirectML/tree/master/Samples/yolov4 - **OpenCL** (Intel, AMD, Mali GPUs for macOS & GNU/Linux) https://github.com/sowson/darknet - **HIP** for Training and Detection on AMD GPU https://github.com/os-hackathon/darknet - **ROS** (Robot Operating System) https://github.com/engcang/ros-yolo-sort - **Xilinx Zynq Ultrascale+ Deep Learning Processor (DPU) ZCU102/ZCU104:** https://github.com/Xilinx/Vitis-In-Depth-Tutorial/tree/master/Machine_Learning/Design_Tutorials/07-yolov4-tutorial - **Amazon Neurochip / Amazon EC2 Inf1 instances** 1.85 times higher throughput and 37% lower cost per image for TensorFlow based YOLOv4 model, using Keras [URL](https://aws.amazon.com/ru/blogs/machine-learning/improving-performance-for-deep-learning-based-object-detection-with-an-aws-neuron-compiled-yolov4-model-on-aws-inferentia/) - **TVM** - compilation of deep learning models (Keras, MXNet, PyTorch, Tensorflow, CoreML, DarkNet) into minimum deployable modules on diverse hardware backend (CPUs, GPUs, FPGA, and specialized accelerators): https://tvm.ai/about - **Tencent/ncnn:** the fastest inference of YOLOv4 on mobile phone CPU: https://github.com/Tencent/ncnn - **OpenDataCam** - It detects, tracks and counts moving objects by using YOLOv4: https://github.com/opendatacam/opendatacam#-hardware-pre-requisite - **Netron** - Visualizer for neural networks: https://github.com/lutzroeder/netron #### Datasets - MS COCO: use `./scripts/get_coco_dataset.sh` to get labeled MS COCO detection dataset - OpenImages: use `python ./scripts/get_openimages_dataset.py` for labeling train detection dataset - Pascal VOC: use `python ./scripts/voc_label.py` for labeling Train/Test/Val detection datasets - ILSVRC2012 (ImageNet classification): use `./scripts/get_imagenet_train.sh` (also `imagenet_label.sh` for labeling valid set) - German/Belgium/Russian/LISA/MASTIF Traffic Sign Datasets for Detection - use this parser: https://github.com/angeligareta/Datasets2Darknet#detection-task - List of other datasets: https://github.com/AlexeyAB/darknet/tree/master/scripts#datasets ### Improvements in this repository - developed State-of-the-Art object detector YOLOv4 - added State-of-Art models: CSP, PRN, EfficientNet - added layers: [conv_lstm], [scale_channels] SE/ASFF/BiFPN, [local_avgpool], [sam], [Gaussian_yolo], [reorg3d] (fixed [reorg]), fixed [batchnorm] - added the ability for training recurrent models (with layers conv-lstm`[conv_lstm]`/conv-rnn`[crnn]`) for accurate detection on video - added data augmentation: `[net] mixup=1 cutmix=1 mosaic=1 blur=1`. Added activations: SWISH, MISH, NORM_CHAN, NORM_CHAN_SOFTMAX - added the ability for training with GPU-processing using CPU-RAM to increase the mini_batch_size and increase accuracy (instead of batch-norm sync) - improved binary neural network performance **2x-4x times** for Detection on CPU and GPU if you trained your own weights by using this XNOR-net model (bit-1 inference) : https://github.com/AlexeyAB/darknet/blob/master/cfg/yolov3-tiny_xnor.cfg - improved neural network performance **~7%** by fusing 2 layers into 1: Convolutional + Batch-norm - improved performance: Detection **2x times**, on GPU Volta/Turing (Tesla V100, GeForce RTX, ...) using Tensor Cores if `CUDNN_HALF` defined in the `Makefile` or `darknet.sln` - improved performance **~1.2x** times on FullHD, **~2x** times on 4K, for detection on the video (file/stream) using `darknet detector demo`... - improved performance **3.5 X times** of data augmentation for training (using OpenCV SSE/AVX functions instead of hand-written functions) - removes bottleneck for training on multi-GPU or GPU Volta - improved performance of detection and training on Intel CPU with AVX (Yolo v3 **~85%**) - optimized memory allocation during network resizing when `random=1` - optimized GPU initialization for detection - we use batch=1 initially instead of re-init with batch=1 - added correct calculation of **mAP, F1, IoU, Precision-Recall** using command `darknet detector map`... - added drawing of chart of average-Loss and accuracy-mAP (`-map` flag) during training - run `./darknet detector demo ... -json_port 8070 -mjpeg_port 8090` as JSON and MJPEG server to get results online over the network by using your soft or Web-browser - added calculation of anchors for training - added example of Detection and Tracking objects: https://github.com/AlexeyAB/darknet/blob/master/src/yolo_console_dll.cpp - run-time tips and warnings if you use incorrect cfg-file or dataset - added support for Windows - many other fixes of code... And added manual - [How to train Yolo v4-v2 (to detect your custom objects)](#how-to-train-to-detect-your-custom-objects) Also, you might be interested in using a simplified repository where is implemented INT8-quantization (+30% speedup and -1% mAP reduced): https://github.com/AlexeyAB/yolo2_light #### How to use on the command line If you use `build.ps1` script or the makefile (Linux only) you will find `darknet` in the root directory. If you use the deprecated Visual Studio solutions, you will find `darknet` in the directory `\build\darknet\x64`. If you customize build with CMake GUI, darknet executable will be installed in your preferred folder. - Yolo v4 COCO - **image**: `./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -thresh 0.25` - **Output coordinates** of objects: `./darknet detector test cfg/coco.data yolov4.cfg yolov4.weights -ext_output dog.jpg` - Yolo v4 COCO - **video**: `./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output test.mp4` - Yolo v4 COCO - **WebCam 0**: `./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -c 0` - Yolo v4 COCO for **net-videocam** - Smart WebCam: `./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights http://192.168.0.80:8080/video?dummy=param.mjpg` - Yolo v4 - **save result videofile res.avi**: `./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights test.mp4 -out_filename res.avi` - Yolo v3 **Tiny** COCO - video: `./darknet detector demo cfg/coco.data cfg/yolov3-tiny.cfg yolov3-tiny.weights test.mp4` - **JSON and MJPEG server** that allows multiple connections from your soft or Web-browser `ip-address:8070` and 8090: `./darknet detector demo ./cfg/coco.data ./cfg/yolov3.cfg ./yolov3.weights test50.mp4 -json_port 8070 -mjpeg_port 8090 -ext_output` - Yolo v3 Tiny **on GPU #1**: `./darknet detector demo cfg/coco.data cfg/yolov3-tiny.cfg yolov3-tiny.weights -i 1 test.mp4` - Alternative method Yolo v3 COCO - image: `./darknet detect cfg/yolov4.cfg yolov4.weights -i 0 -thresh 0.25` - Train on **Amazon EC2**, to see mAP & Loss-chart using URL like: `http://ec2-35-160-228-91.us-west-2.compute.amazonaws.com:8090` in the Chrome/Firefox (**Darknet should be compiled with OpenCV**): `./darknet detector train cfg/coco.data yolov4.cfg yolov4.conv.137 -dont_show -mjpeg_port 8090 -map` - 186 MB Yolo9000 - image: `./darknet detector test cfg/combine9k.data cfg/yolo9000.cfg yolo9000.weights` - Remember to put data/9k.tree and data/coco9k.map under the same folder of your app if you use the cpp api to build an app - To process a list of images `data/train.txt` and save results of detection to `result.json` file use: `./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < data/train.txt` - To process a list of images `data/train.txt` and save results of detection to `result.txt` use: `./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -dont_show -ext_output < data/train.txt > result.txt` - To process a video and output results to a json file use: `darknet.exe detector demo cfg/coco.data cfg/yolov3.cfg yolov3.weights file.mp4 -dont_show -json_file_output results.json` - Pseudo-labelling - to process a list of images `data/new_train.txt` and save results of detection in Yolo training format for each image as label `.txt` (in this way you can increase the amount of training data) use: `./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -thresh 0.25 -dont_show -save_labels < data/new_train.txt` - To calculate anchors: `./darknet detector calc_anchors data/obj.data -num_of_clusters 9 -width 416 -height 416` - To check accuracy mAP@IoU=50: `./darknet detector map data/obj.data yolo-obj.cfg backup\yolo-obj_7000.weights` - To check accuracy mAP@IoU=75: `./darknet detector map data/obj.data yolo-obj.cfg backup\yolo-obj_7000.weights -iou_thresh 0.75` ##### For using network video-camera mjpeg-stream with any Android smartphone 1. Download for Android phone mjpeg-stream soft: IP Webcam / Smart WebCam - Smart WebCam - preferably: https://play.google.com/store/apps/details?id=com.acontech.android.SmartWebCam2 - IP Webcam: https://play.google.com/store/apps/details?id=com.pas.webcam 2. Connect your Android phone to the computer by WiFi (through a WiFi-router) or USB 3. Start Smart WebCam on your phone 4. Replace the address below, shown in the phone application (Smart WebCam) and launch: - Yolo v4 COCO-model: `./darknet detector demo data/coco.data yolov4.cfg yolov4.weights http://192.168.0.80:8080/video?dummy=param.mjpg -i 0` ### How to compile on Linux/macOS (using `CMake`) The `CMakeLists.txt` will attempt to find installed optional dependencies like CUDA, cudnn, ZED and build against those. It will also create a shared object library file to use `darknet` for code development. To update CMake on Ubuntu, it's better to follow guide here: https://apt.kitware.com/ or https://cmake.org/download/ ```bash git clone https://github.com/AlexeyAB/darknet cd darknet mkdir build_release cd build_release cmake .. cmake --build . --target install --parallel 8 ``` ### Using also PowerShell Install: `Cmake`, `CUDA`, `cuDNN` [How to install dependencies](#requirements) Install powershell for your OS (Linux or MacOS) ([guide here](https://docs.microsoft.com/en-us/powershell/scripting/install/installing-powershell)). Open PowerShell type these commands ```PowerShell git clone https://github.com/AlexeyAB/darknet cd darknet ./build.ps1 -UseVCPKG -EnableOPENCV -EnableCUDA -EnableCUDNN ``` - remove options like `-EnableCUDA` or `-EnableCUDNN` if you are not interested into - remove option `-UseVCPKG` if you plan to manually provide OpenCV library to darknet or if you do not want to enable OpenCV integration - add option `-EnableOPENCV_CUDA` if you want to build OpenCV with CUDA support - very slow to build! (requires `-UseVCPKG`) If you open the `build.ps1` script at the beginning you will find all available switches. ### How to compile on Linux (using `make`) Just do `make` in the darknet directory. (You can try to compile and run it on Google Colab in cloud [link](https://colab.research.google.com/drive/12QusaaRj_lUwCGDvQNfICpa7kA7_a2dE) (press «Open in Playground» button at the top-left corner) and watch the video [link](https://www.youtube.com/watch?v=mKAEGSxwOAY) ) Before make, you can set such options in the `Makefile`: [link](https://github.com/AlexeyAB/darknet/blob/9c1b9a2cf6363546c152251be578a21f3c3caec6/Makefile#L1) - `GPU=1` to build with CUDA to accelerate by using GPU (CUDA should be in `/usr/local/cuda`) - `CUDNN=1` to build with cuDNN v5-v7 to accelerate training by using GPU (cuDNN should be in `/usr/local/cudnn`) - `CUDNN_HALF=1` to build for Tensor Cores (on Titan V / Tesla V100 / DGX-2 and later) speedup Detection 3x, Training 2x - `OPENCV=1` to build with OpenCV 4.x/3.x/2.4.x - allows to detect on video files and video streams from network cameras or web-cams - `DEBUG=1` to build debug version of Yolo - `OPENMP=1` to build with OpenMP support to accelerate Yolo by using multi-core CPU - `LIBSO=1` to build a library `darknet.so` and binary runnable file `uselib` that uses this library. Or you can try to run so `LD_LIBRARY_PATH=./:$LD_LIBRARY_PATH ./uselib test.mp4` How to use this SO-library from your own code - you can look at C++ example: https://github.com/AlexeyAB/darknet/blob/master/src/yolo_console_dll.cpp or use in such a way: `LD_LIBRARY_PATH=./:$LD_LIBRARY_PATH ./uselib data/coco.names cfg/yolov4.cfg yolov4.weights test.mp4` - `ZED_CAMERA=1` to build a library with ZED-3D-camera support (should be ZED SDK installed), then run `LD_LIBRARY_PATH=./:$LD_LIBRARY_PATH ./uselib data/coco.names cfg/yolov4.cfg yolov4.weights zed_camera` - You also need to specify for which graphics card the code is generated. This is done by setting `ARCH=`. If you use a newer version than CUDA 11 you further need to edit line 20 from Makefile and remove `-gencode arch=compute_30,code=sm_30 \` as Kepler GPU support was dropped in CUDA 11. You can also drop the general `ARCH=` and just uncomment `ARCH=` for your graphics card. ### How to compile on Windows (using `CMake`) Requires: - MSVC: https://visualstudio.microsoft.com/thank-you-downloading-visual-studio/?sku=Community - CMake GUI: `Windows win64-x64 Installer`https://cmake.org/download/ - Download Darknet zip-archive with the latest commit and uncompress it: [master.zip](https://github.com/AlexeyAB/darknet/archive/master.zip) In Windows: - Start (button) -> All programs -> CMake -> CMake (gui) -> - [look at image](https://habrastorage.org/webt/pz/s1/uu/pzs1uu4heb7vflfcjqn-lxy-aqu.jpeg) In CMake: Enter input path to the darknet Source, and output path to the Binaries -> Configure (button) -> Optional platform for generator: `x64` -> Finish -> Generate -> Open Project -> - in MS Visual Studio: Select: x64 and Release -> Build -> Build solution - find the executable file `darknet.exe` in the output path to the binaries you specified ![x64 and Release](https://habrastorage.org/webt/ay/ty/f-/aytyf-8bufe7q-16yoecommlwys.jpeg) ### How to compile on Windows (using `vcpkg`) This is the recommended approach to build Darknet on Windows. 1. Install Visual Studio 2017 or 2019. In case you need to download it, please go here: [Visual Studio Community](http://visualstudio.com). Remember to install English language pack, this is mandatory for vcpkg! 2. Install CUDA enabling VS Integration during installation. 3. Open Powershell (Start -> All programs -> Windows Powershell) and type these commands: ```PowerShell Set-ExecutionPolicy unrestricted -Scope CurrentUser -Force git clone https://github.com/AlexeyAB/darknet cd darknet .\build.ps1 -UseVCPKG -EnableOPENCV -EnableCUDA -EnableCUDNN ``` (add option `-EnableOPENCV_CUDA` if you want to build OpenCV with CUDA support - very slow to build! - or remove options like `-EnableCUDA` or `-EnableCUDNN` if you are not interested in them). If you open the `build.ps1` script at the beginning you will find all available switches. ## How to train with multi-GPU 1. Train it first on 1 GPU for like 1000 iterations: `darknet.exe detector train cfg/coco.data cfg/yolov4.cfg yolov4.conv.137` 2. Then stop and by using partially-trained model `/backup/yolov4_1000.weights` run training with multigpu (up to 4 GPUs): `darknet.exe detector train cfg/coco.data cfg/yolov4.cfg /backup/yolov4_1000.weights -gpus 0,1,2,3` If you get a Nan, then for some datasets better to decrease learning rate, for 4 GPUs set `learning_rate = 0,00065` (i.e. learning_rate = 0.00261 / GPUs). In this case also increase 4x times `burn_in =` in your cfg-file. I.e. use `burn_in = 4000` instead of `1000`. https://groups.google.com/d/msg/darknet/NbJqonJBTSY/Te5PfIpuCAAJ ## How to train (to detect your custom objects) (to train old Yolo v2 `yolov2-voc.cfg`, `yolov2-tiny-voc.cfg`, `yolo-voc.cfg`, `yolo-voc.2.0.cfg`, ... [click by the link](https://github.com/AlexeyAB/darknet/tree/47c7af1cea5bbdedf1184963355e6418cb8b1b4f#how-to-train-pascal-voc-data)) Training Yolo v4 (and v3): 0. For training `cfg/yolov4-custom.cfg` download the pre-trained weights-file (162 MB): [yolov4.conv.137](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.conv.137) (Google drive mirror [yolov4.conv.137](https://drive.google.com/open?id=1JKF-bdIklxOOVy-2Cr5qdvjgGpmGfcbp) ) 1. Create file `yolo-obj.cfg` with the same content as in `yolov4-custom.cfg` (or copy `yolov4-custom.cfg` to `yolo-obj.cfg)` and: - change line batch to [`batch=64`](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L3) - change line subdivisions to [`subdivisions=16`](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L4) - change line max_batches to (`classes*2000`, but not less than number of training images and not less than `6000`), f.e. [`max_batches=6000`](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L20) if you train for 3 classes - change line steps to 80% and 90% of max_batches, f.e. [`steps=4800,5400`](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L22) - set network size `width=416 height=416` or any value multiple of 32: https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L8-L9 - change line `classes=80` to your number of objects in each of 3 `[yolo]`-layers: - https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L610 - https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L696 - https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L783 - change [`filters=255`] to filters=(classes + 5)x3 in the 3 `[convolutional]` before each `[yolo]` layer, keep in mind that it only has to be the last `[convolutional]` before each of the `[yolo]` layers. - https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L603 - https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L689 - https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L776 - when using [`[Gaussian_yolo]`](https://github.com/AlexeyAB/darknet/blob/6e5bdf1282ad6b06ed0e962c3f5be67cf63d96dc/cfg/Gaussian_yolov3_BDD.cfg#L608) layers, change [`filters=57`] filters=(classes + 9)x3 in the 3 `[convolutional]` before each `[Gaussian_yolo]` layer - https://github.com/AlexeyAB/darknet/blob/6e5bdf1282ad6b06ed0e962c3f5be67cf63d96dc/cfg/Gaussian_yolov3_BDD.cfg#L604 - https://github.com/AlexeyAB/darknet/blob/6e5bdf1282ad6b06ed0e962c3f5be67cf63d96dc/cfg/Gaussian_yolov3_BDD.cfg#L696 - https://github.com/AlexeyAB/darknet/blob/6e5bdf1282ad6b06ed0e962c3f5be67cf63d96dc/cfg/Gaussian_yolov3_BDD.cfg#L789 So if `classes=1` then should be `filters=18`. If `classes=2` then write `filters=21`. **(Do not write in the cfg-file: filters=(classes + 5)x3)** (Generally `filters` depends on the `classes`, `coords` and number of `mask`s, i.e. filters=`(classes + coords + 1)*`, where `mask` is indices of anchors. If `mask` is absence, then filters=`(classes + coords + 1)*num`) So for example, for 2 objects, your file `yolo-obj.cfg` should differ from `yolov4-custom.cfg` in such lines in each of **3** [yolo]-layers: ```ini [convolutional] filters=21 [region] classes=2 ``` 2. Create file `obj.names` in the directory `build\darknet\x64\data\`, with objects names - each in new line 3. Create file `obj.data` in the directory `build\darknet\x64\data\`, containing (where **classes = number of objects**): ```ini classes = 2 train = data/train.txt valid = data/test.txt names = data/obj.names backup = backup/ ``` 4. Put image-files (.jpg) of your objects in the directory `build\darknet\x64\data\obj\` 5. You should label each object on images from your dataset. Use this visual GUI-software for marking bounded boxes of objects and generating annotation files for Yolo v2 & v3: https://github.com/AlexeyAB/Yolo_mark It will create `.txt`-file for each `.jpg`-image-file - in the same directory and with the same name, but with `.txt`-extension, and put to file: object number and object coordinates on this image, for each object in new line: ` ` Where: - `` - integer object number from `0` to `(classes-1)` - ` ` - float values **relative** to width and height of image, it can be equal from `(0.0 to 1.0]` - for example: ` = / ` or ` = / ` - attention: ` ` - are center of rectangle (are not top-left corner) For example for `img1.jpg` you will be created `img1.txt` containing: ```csv 1 0.716797 0.395833 0.216406 0.147222 0 0.687109 0.379167 0.255469 0.158333 1 0.420312 0.395833 0.140625 0.166667 ``` 6. Create file `train.txt` in directory `build\darknet\x64\data\`, with filenames of your images, each filename in new line, with path relative to `darknet.exe`, for example containing: ```csv data/obj/img1.jpg data/obj/img2.jpg data/obj/img3.jpg ``` 7. Download pre-trained weights for the convolutional layers and put to the directory `build\darknet\x64` - for `yolov4.cfg`, `yolov4-custom.cfg` (162 MB): [yolov4.conv.137](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.conv.137) (Google drive mirror [yolov4.conv.137](https://drive.google.com/open?id=1JKF-bdIklxOOVy-2Cr5qdvjgGpmGfcbp) ) - for `yolov4-tiny.cfg`, `yolov4-tiny-3l.cfg`, `yolov4-tiny-custom.cfg` (19 MB): [yolov4-tiny.conv.29](https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.conv.29) - for `csresnext50-panet-spp.cfg` (133 MB): [csresnext50-panet-spp.conv.112](https://drive.google.com/file/d/16yMYCLQTY_oDlCIZPfn_sab6KD3zgzGq/view?usp=sharing) - for `yolov3.cfg, yolov3-spp.cfg` (154 MB): [darknet53.conv.74](https://pjreddie.com/media/files/darknet53.conv.74) - for `yolov3-tiny-prn.cfg , yolov3-tiny.cfg` (6 MB): [yolov3-tiny.conv.11](https://drive.google.com/file/d/18v36esoXCh-PsOKwyP2GWrpYDptDY8Zf/view?usp=sharing) - for `enet-coco.cfg (EfficientNetB0-Yolov3)` (14 MB): [enetb0-coco.conv.132](https://drive.google.com/file/d/1uhh3D6RSn0ekgmsaTcl-ZW53WBaUDo6j/view?usp=sharing) 8. Start training by using the command line: `darknet.exe detector train data/obj.data yolo-obj.cfg yolov4.conv.137` To train on Linux use command: `./darknet detector train data/obj.data yolo-obj.cfg yolov4.conv.137` (just use `./darknet` instead of `darknet.exe`) - (file `yolo-obj_last.weights` will be saved to the `build\darknet\x64\backup\` for each 100 iterations) - (file `yolo-obj_xxxx.weights` will be saved to the `build\darknet\x64\backup\` for each 1000 iterations) - (to disable Loss-Window use `darknet.exe detector train data/obj.data yolo-obj.cfg yolov4.conv.137 -dont_show`, if you train on computer without monitor like a cloud Amazon EC2) - (to see the mAP & Loss-chart during training on remote server without GUI, use command `darknet.exe detector train data/obj.data yolo-obj.cfg yolov4.conv.137 -dont_show -mjpeg_port 8090 -map` then open URL `http://ip-address:8090` in Chrome/Firefox browser) 8.1. For training with mAP (mean average precisions) calculation for each 4 Epochs (set `valid=valid.txt` or `train.txt` in `obj.data` file) and run: `darknet.exe detector train data/obj.data yolo-obj.cfg yolov4.conv.137 -map` 8.2. One can also set the `-mAP_epochs` in the training command if less or more frequent mAP calculation is needed. For example in order to calculate mAP for each 2 Epochs run `darknet.exe detector train data/obj.data yolo-obj.cfg yolov4.conv.137 -map -mAP_epochs 2` 9. After training is complete - get result `yolo-obj_final.weights` from path `build\darknet\x64\backup\` - After each 100 iterations you can stop and later start training from this point. For example, after 2000 iterations you can stop training, and later just start training using: `darknet.exe detector train data/obj.data yolo-obj.cfg backup\yolo-obj_2000.weights` (in the original repository https://github.com/pjreddie/darknet the weights-file is saved only once every 10 000 iterations `if(iterations > 1000)`) - Also you can get result earlier than all 45000 iterations. **Note:** If during training you see `nan` values for `avg` (loss) field - then training goes wrong, but if `nan` is in some other lines - then training goes well. **Note:** If you changed width= or height= in your cfg-file, then new width and height must be divisible by 32. **Note:** After training use such command for detection: `darknet.exe detector test data/obj.data yolo-obj.cfg yolo-obj_8000.weights` **Note:** if error `Out of memory` occurs then in `.cfg`-file you should increase `subdivisions=16`, 32 or 64: [link](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L4) ### How to train tiny-yolo (to detect your custom objects) Do all the same steps as for the full yolo model as described above. With the exception of: - Download file with the first 29-convolutional layers of yolov4-tiny: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v4_pre/yolov4-tiny.conv.29 (Or get this file from yolov4-tiny.weights file by using command: `darknet.exe partial cfg/yolov4-tiny-custom.cfg yolov4-tiny.weights yolov4-tiny.conv.29 29` - Make your custom model `yolov4-tiny-obj.cfg` based on `cfg/yolov4-tiny-custom.cfg` instead of `yolov4.cfg` - Start training: `darknet.exe detector train data/obj.data yolov4-tiny-obj.cfg yolov4-tiny.conv.29` For training Yolo based on other models ([DenseNet201-Yolo](https://github.com/AlexeyAB/darknet/blob/master/build/darknet/x64/densenet201_yolo.cfg) or [ResNet50-Yolo](https://github.com/AlexeyAB/darknet/blob/master/build/darknet/x64/resnet50_yolo.cfg)), you can download and get pre-trained weights as showed in this file: https://github.com/AlexeyAB/darknet/blob/master/build/darknet/x64/partial.cmd If you made you custom model that isn't based on other models, then you can train it without pre-trained weights, then will be used random initial weights. ## When should I stop training Usually sufficient 2000 iterations for each class(object), but not less than number of training images and not less than 6000 iterations in total. But for a more precise definition of when you should stop training, use the following manual: 1. During training, you will see varying indicators of error, and you should stop when no longer decreases **0.XXXXXXX avg**: > Region Avg IOU: 0.798363, Class: 0.893232, Obj: 0.700808, No Obj: 0.004567, Avg Recall: 1.000000, count: 8 > Region Avg IOU: 0.800677, Class: 0.892181, Obj: 0.701590, No Obj: 0.004574, Avg Recall: 1.000000, count: 8 > > **9002**: 0.211667, **0.60730 avg**, 0.001000 rate, 3.868000 seconds, 576128 images > Loaded: 0.000000 seconds - **9002** - iteration number (number of batch) - **0.60730 avg** - average loss (error) - **the lower, the better** When you see that average loss **0.xxxxxx avg** no longer decreases at many iterations then you should stop training. The final average loss can be from `0.05` (for a small model and easy dataset) to `3.0` (for a big model and a difficult dataset). Or if you train with flag `-map` then you will see mAP indicator `Last accuracy mAP@0.5 = 18.50%` in the console - this indicator is better than Loss, so train while mAP increases. 2. Once training is stopped, you should take some of last `.weights`-files from `darknet\build\darknet\x64\backup` and choose the best of them: For example, you stopped training after 9000 iterations, but the best result can give one of previous weights (7000, 8000, 9000). It can happen due to over-fitting. **Over-fitting** - is case when you can detect objects on images from training-dataset, but can't detect objects on any others images. You should get weights from **Early Stopping Point**: ![Over-fitting](https://hsto.org/files/5dc/7ae/7fa/5dc7ae7fad9d4e3eb3a484c58bfc1ff5.png) To get weights from Early Stopping Point: 2.1. At first, in your file `obj.data` you must specify the path to the validation dataset `valid = valid.txt` (format of `valid.txt` as in `train.txt`), and if you haven't validation images, just copy `data\train.txt` to `data\valid.txt`. 2.2 If training is stopped after 9000 iterations, to validate some of previous weights use this commands: (If you use another GitHub repository, then use `darknet.exe detector recall`... instead of `darknet.exe detector map`...) - `darknet.exe detector map data/obj.data yolo-obj.cfg backup\yolo-obj_7000.weights` - `darknet.exe detector map data/obj.data yolo-obj.cfg backup\yolo-obj_8000.weights` - `darknet.exe detector map data/obj.data yolo-obj.cfg backup\yolo-obj_9000.weights` And compare last output lines for each weights (7000, 8000, 9000): Choose weights-file **with the highest mAP (mean average precision)** or IoU (intersect over union) For example, **bigger mAP** gives weights `yolo-obj_8000.weights` - then **use this weights for detection**. Or just train with `-map` flag: `darknet.exe detector train data/obj.data yolo-obj.cfg yolov4.conv.137 -map` So you will see mAP-chart (red-line) in the Loss-chart Window. mAP will be calculated for each 4 Epochs using `valid=valid.txt` file that is specified in `obj.data` file (`1 Epoch = images_in_train_txt / batch` iterations) (to change the max x-axis value - change [`max_batches=`](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L20) parameter to `2000*classes`, f.e. `max_batches=6000` for 3 classes) ![loss_chart_map_chart](https://hsto.org/webt/yd/vl/ag/ydvlagutof2zcnjodstgroen8ac.jpeg) Example of custom object detection: `darknet.exe detector test data/obj.data yolo-obj.cfg yolo-obj_8000.weights` - **IoU** (intersect over union) - average intersect over union of objects and detections for a certain threshold = 0.24 - **mAP** (mean average precision) - mean value of `average precisions` for each class, where `average precision` is average value of 11 points on PR-curve for each possible threshold (each probability of detection) for the same class (Precision-Recall in terms of PascalVOC, where Precision=TP/(TP+FP) and Recall=TP/(TP+FN) ), page-11: http://homepages.inf.ed.ac.uk/ckiw/postscript/ijcv_voc09.pdf **mAP** is default metric of precision in the PascalVOC competition, **this is the same as AP50** metric in the MS COCO competition. In terms of Wiki, indicators Precision and Recall have a slightly different meaning than in the PascalVOC competition, but **IoU always has the same meaning**. ![precision_recall_iou](https://hsto.org/files/ca8/866/d76/ca8866d76fb840228940dbf442a7f06a.jpg) ### Custom object detection Example of custom object detection: `darknet.exe detector test data/obj.data yolo-obj.cfg yolo-obj_8000.weights` | ![Yolo_v2_training](https://hsto.org/files/d12/1e7/515/d121e7515f6a4eb694913f10de5f2b61.jpg) | ![Yolo_v2_training](https://hsto.org/files/727/c7e/5e9/727c7e5e99bf4d4aa34027bb6a5e4bab.jpg) | |---|---| ## How to improve object detection 1. Before training: - set flag `random=1` in your `.cfg`-file - it will increase precision by training Yolo for different resolutions: [link](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L788) - increase network resolution in your `.cfg`-file (`height=608`, `width=608` or any value multiple of 32) - it will increase precision - check that each object that you want to detect is mandatory labeled in your dataset - no one object in your data set should not be without label. In the most training issues - there are wrong labels in your dataset (got labels by using some conversion script, marked with a third-party tool, ...). Always check your dataset by using: https://github.com/AlexeyAB/Yolo_mark - my Loss is very high and mAP is very low, is training wrong? Run training with `-show_imgs` flag at the end of training command, do you see correct bounded boxes of objects (in windows or in files `aug_...jpg`)? If no - your training dataset is wrong. - for each object which you want to detect - there must be at least 1 similar object in the Training dataset with about the same: shape, side of object, relative size, angle of rotation, tilt, illumination. So desirable that your training dataset include images with objects at different: scales, rotations, lightings, from different sides, on different backgrounds - you should preferably have 2000 different images for each class or more, and you should train `2000*classes` iterations or more - desirable that your training dataset include images with non-labeled objects that you do not want to detect - negative samples without bounded box (empty `.txt` files) - use as many images of negative samples as there are images with objects - What is the best way to mark objects: label only the visible part of the object, or label the visible and overlapped part of the object, or label a little more than the entire object (with a little gap)? Mark as you like - how would you like it to be detected. - for training with a large number of objects in each image, add the parameter `max=200` or higher value in the last `[yolo]`-layer or `[region]`-layer in your cfg-file (the global maximum number of objects that can be detected by YoloV3 is `0,0615234375*(width*height)` where are width and height are parameters from `[net]` section in cfg-file) - for training for small objects (smaller than 16x16 after the image is resized to 416x416) - set `layers = 23` instead of https://github.com/AlexeyAB/darknet/blob/6f718c257815a984253346bba8fb7aa756c55090/cfg/yolov4.cfg#L895 - set `stride=4` instead of https://github.com/AlexeyAB/darknet/blob/6f718c257815a984253346bba8fb7aa756c55090/cfg/yolov4.cfg#L892 - set `stride=4` instead of https://github.com/AlexeyAB/darknet/blob/6f718c257815a984253346bba8fb7aa756c55090/cfg/yolov4.cfg#L989 - for training for both small and large objects use modified models: - Full-model: 5 yolo layers: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov3_5l.cfg - Tiny-model: 3 yolo layers: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-tiny-3l.cfg - YOLOv4: 3 yolo layers: https://raw.githubusercontent.com/AlexeyAB/darknet/master/cfg/yolov4-custom.cfg - If you train the model to distinguish Left and Right objects as separate classes (left/right hand, left/right-turn on road signs, ...) then for disabling flip data augmentation - add `flip=0` here: https://github.com/AlexeyAB/darknet/blob/3d2d0a7c98dbc8923d9ff705b81ff4f7940ea6ff/cfg/yolov3.cfg#L17 - General rule - your training dataset should include such a set of relative sizes of objects that you want to detect: - `train_network_width * train_obj_width / train_image_width ~= detection_network_width * detection_obj_width / detection_image_width` - `train_network_height * train_obj_height / train_image_height ~= detection_network_height * detection_obj_height / detection_image_height` I.e. for each object from Test dataset there must be at least 1 object in the Training dataset with the same class_id and about the same relative size: `object width in percent from Training dataset` ~= `object width in percent from Test dataset` That is, if only objects that occupied 80-90% of the image were present in the training set, then the trained network will not be able to detect objects that occupy 1-10% of the image. - to speedup training (with decreasing detection accuracy) set param `stopbackward=1` for layer-136 in cfg-file - each: `model of object, side, illumination, scale, each 30 grad` of the turn and inclination angles - these are *different objects* from an internal perspective of the neural network. So the more *different objects* you want to detect, the more complex network model should be used. - to make the detected bounded boxes more accurate, you can add 3 parameters `ignore_thresh = .9 iou_normalizer=0.5 iou_loss=giou` to each `[yolo]` layer and train, it will increase mAP@0.9, but decrease mAP@0.5. - Only if you are an **expert** in neural detection networks - recalculate anchors for your dataset for `width` and `height` from cfg-file: `darknet.exe detector calc_anchors data/obj.data -num_of_clusters 9 -width 416 -height 416` then set the same 9 `anchors` in each of 3 `[yolo]`-layers in your cfg-file. But you should change indexes of anchors `masks=` for each [yolo]-layer, so for YOLOv4 the 1st-[yolo]-layer has anchors smaller than 30x30, 2nd smaller than 60x60, 3rd remaining, and vice versa for YOLOv3. Also you should change the `filters=(classes + 5)*` before each [yolo]-layer. If many of the calculated anchors do not fit under the appropriate layers - then just try using all the default anchors. 2. After training - for detection: - Increase network-resolution by set in your `.cfg`-file (`height=608` and `width=608`) or (`height=832` and `width=832`) or (any value multiple of 32) - this increases the precision and makes it possible to detect small objects: [link](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L8-L9) - it is not necessary to train the network again, just use `.weights`-file already trained for 416x416 resolution - to get even greater accuracy you should train with higher resolution 608x608 or 832x832, note: if error `Out of memory` occurs then in `.cfg`-file you should increase `subdivisions=16`, 32 or 64: [link](https://github.com/AlexeyAB/darknet/blob/0039fd26786ab5f71d5af725fc18b3f521e7acfd/cfg/yolov3.cfg#L4) ## How to mark bounded boxes of objects and create annotation files Here you can find repository with GUI-software for marking bounded boxes of objects and generating annotation files for Yolo v2 - v4: https://github.com/AlexeyAB/Yolo_mark With example of: `train.txt`, `obj.names`, `obj.data`, `yolo-obj.cfg`, `air`1-6`.txt`, `bird`1-4`.txt` for 2 classes of objects (air, bird) and `train_obj.cmd` with example how to train this image-set with Yolo v2 - v4 Different tools for marking objects in images: 1. in C++: https://github.com/AlexeyAB/Yolo_mark 2. in Python: https://github.com/tzutalin/labelImg 3. in Python: https://github.com/Cartucho/OpenLabeling 4. in C++: https://www.ccoderun.ca/darkmark/ 5. in JavaScript: https://github.com/opencv/cvat 6. in C++: https://github.com/jveitchmichaelis/deeplabel 7. in C#: https://github.com/BMW-InnovationLab/BMW-Labeltool-Lite 8. DL-Annotator for Windows ($30): [url](https://www.microsoft.com/en-us/p/dlannotator/9nsx79m7t8fn?activetab=pivot:overviewtab) 9. v7labs - the greatest cloud labeling tool ($1.5 per hour): https://www.v7labs.com/ ## How to use Yolo as DLL and SO libraries - on Linux - using `build.sh` or - build `darknet` using `cmake` or - set `LIBSO=1` in the `Makefile` and do `make` - on Windows - using `build.ps1` or - build `darknet` using `cmake` or - compile `build\darknet\yolo_cpp_dll.sln` solution or `build\darknet\yolo_cpp_dll_no_gpu.sln` solution There are 2 APIs: - C API: https://github.com/AlexeyAB/darknet/blob/master/include/darknet.h - Python examples using the C API: - https://github.com/AlexeyAB/darknet/blob/master/darknet.py - https://github.com/AlexeyAB/darknet/blob/master/darknet_video.py - C++ API: https://github.com/AlexeyAB/darknet/blob/master/include/yolo_v2_class.hpp - C++ example that uses C++ API: https://github.com/AlexeyAB/darknet/blob/master/src/yolo_console_dll.cpp ---- 1. To compile Yolo as C++ DLL-file `yolo_cpp_dll.dll` - open the solution `build\darknet\yolo_cpp_dll.sln`, set **x64** and **Release**, and do the: Build -> Build yolo_cpp_dll - You should have installed **CUDA 10.2** - To use cuDNN do: (right click on project) -> properties -> C/C++ -> Preprocessor -> Preprocessor Definitions, and add at the beginning of line: `CUDNN;` 2. To use Yolo as DLL-file in your C++ console application - open the solution `build\darknet\yolo_console_dll.sln`, set **x64** and **Release**, and do the: Build -> Build yolo_console_dll - you can run your console application from Windows Explorer `build\darknet\x64\yolo_console_dll.exe` **use this command**: `yolo_console_dll.exe data/coco.names yolov4.cfg yolov4.weights test.mp4` - after launching your console application and entering the image file name - you will see info for each object: ` ` - to use simple OpenCV-GUI you should uncomment line `//#define OPENCV` in `yolo_console_dll.cpp`-file: [link](https://github.com/AlexeyAB/darknet/blob/a6cbaeecde40f91ddc3ea09aa26a03ab5bbf8ba8/src/yolo_console_dll.cpp#L5) - you can see source code of simple example for detection on the video file: [link](https://github.com/AlexeyAB/darknet/blob/ab1c5f9e57b4175f29a6ef39e7e68987d3e98704/src/yolo_console_dll.cpp#L75) `yolo_cpp_dll.dll`-API: [link](https://github.com/AlexeyAB/darknet/blob/master/src/yolo_v2_class.hpp#L42) ```cpp struct bbox_t { unsigned int x, y, w, h; // (x,y) - top-left corner, (w, h) - width & height of bounded box float prob; // confidence - probability that the object was found correctly unsigned int obj_id; // class of object - from range [0, classes-1] unsigned int track_id; // tracking id for video (0 - untracked, 1 - inf - tracked object) unsigned int frames_counter;// counter of frames on which the object was detected }; class Detector { public: Detector(std::string cfg_filename, std::string weight_filename, int gpu_id = 0); ~Detector(); std::vector detect(std::string image_filename, float thresh = 0.2, bool use_mean = false); std::vector detect(image_t img, float thresh = 0.2, bool use_mean = false); static image_t load_image(std::string image_filename); static void free_image(image_t m); #ifdef OPENCV std::vector detect(cv::Mat mat, float thresh = 0.2, bool use_mean = false); std::shared_ptr mat_to_image_resize(cv::Mat mat) const; #endif }; ``` ## Citation ``` @misc{bochkovskiy2020yolov4, title={YOLOv4: Optimal Speed and Accuracy of Object Detection}, author={Alexey Bochkovskiy and Chien-Yao Wang and Hong-Yuan Mark Liao}, year={2020}, eprint={2004.10934}, archivePrefix={arXiv}, primaryClass={cs.CV} } ``` ``` @InProceedings{Wang_2021_CVPR, author = {Wang, Chien-Yao and Bochkovskiy, Alexey and Liao, Hong-Yuan Mark}, title = {{Scaled-YOLOv4}: Scaling Cross Stage Partial Network}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, month = {June}, year = {2021}, pages = {13029-13038} } ```