Skip to content

[yolov5 v7.0][WIDER FACE][ECCV 2022]YOLO5Face: Why Reinventing a Face Detector

License

Notifications You must be signed in to change notification settings

zjykzj/YOLO5Face

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

96 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Language: πŸ‡ΊπŸ‡Έ πŸ‡¨πŸ‡³

Β«YOLO5FaceΒ» reproduced the paper "YOLO5Face: Why Reinventing a Face Detector"

ONLINE DEMO:FACE AND LANDMARKS DETECT

ARCH GFLOPs Easy Medium Hard
zjykzj/YOLO5Face (This) yolov5s-face 15.2 94.69 93.00 84.73
deepcam-cn/yolov5-face(Official) yolov5s-face / 94.33 92.61 83.15
zjykzj/YOLO5Face (This) shufflenetv2-face 1.5 90.27 87.39 73.60
deepcam-cn/yolov5-face(Official) shufflenetv2-face / 90.76 88.12 73.82
zjykzj/YOLO5Face (This) yolov5x-v7.0 204 95.79 94.53 87.63
zjykzj/YOLO5Face (This) yolov5s-v7.0 15.8 94.84 93.28 84.67
zjykzj/YOLO5Face (This) yolov5n-v7.0 4.2 93.25 91.11 80.33

Table of Contents

NewsπŸš€πŸš€πŸš€

Version Release Date Major Updates
v1.1.1 2024/08/17 Update and add Gradio demo.
v1.1.0 2024/07/21 Supports additional models including shufflenetv2-face/yolov5x-v7.0/yolov5n-v7.0.
v1.0.0 2024/07/14 Adds keypoint detection, enabling face + keypoint detection.
v0.1.0 2024/06/29 Trains a face detector based on yolov5-v7.0 and the WIDERFACE dataset.

Background

YOLO5Face is a very interesting work that further abstracts the task of face detection. By directly using a universal object detection algorithm, good face detection results can be achieved. Of course, it also achieves 5-point facial keypoint regression. Based on ultralytics/yolov5, YOLO5Face can easily apply different models and training, such as using lightweight networks for real-time detection and using large networks for higher detection accuracy.

Note: the latest implementation of YOLO5Face in our warehouse is entirely based on ultralytics/yolov5 v7.0.

Installation

$ pip3 install -r requirements.txt

Or use docker container

$ docker run -it --runtime nvidia --gpus=all --shm-size=16g -v /etc/localtime:/etc/localtime -v $(pwd):/workdir --workdir=/workdir --name yolo5face ultralytics/yolov5:v7.0

Usage

Train

# yolov5s_v7.0
$ python3 widerface_train.py --data widerface.yaml --weights "" --cfg models/yolo5face/cfgs/yolov5s_v7_0.yaml --hyp models/yolo5face/hyps/hyp.scratch-low.yaml --img 800 --epoch 300 --device 0
# yolov5s-face
$ python3 widerface_train.py --data widerface.yaml --weights "" --cfg models/yolo5face/cfgs/yolov5s_face.yaml --hyp models/yolo5face/hyps/hyp.scratch.yaml --img 800 --epoch 300 --device 0

Eval

$ python widerface_detect.py --weights ./runs/exp4-yolov5s_v7_0-i800-e300.pt --source ../datasets/widerface/images/val/ --folder_pict ../datasets/widerface/wider_face_split/wider_face_val_bbx_gt.txt --conf-thres 0.001 --iou-thres 0.6 --save-txt --save-conf --device 0
...
YOLOv5s_v7_0 summary: 157 layers, 7039792 parameters, 0 gradients, 15.8 GFLOPs
...
Speed: 0.4ms pre-process, 8.8ms inference, 0.8ms NMS per image at shape (1, 3, 640, 640)
Results saved to runs/detect/exp7
0 labels saved to runs/detect/exp7/labels
$ cd widerface_evaluate/
$ python3 evaluation.py -p ../runs/detect/exp7/labels/ -g ./ground_truth/
Reading Predictions : 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 61/61 [00:00<00:00, 62.18it/s]
Processing easy: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 61/61 [00:20<00:00,  2.94it/s]
Processing medium: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 61/61 [00:20<00:00,  2.98it/s]
Processing hard: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 61/61 [00:20<00:00,  2.97it/s]
==================== Results ====================
Easy   Val AP: 0.9483604102331251
Medium Val AP: 0.9328484206418773
Hard   Val AP: 0.8467345083774318
=================================================

Predict

$ python detect_face_and_landmarks.py --weights ./runs/exp4-yolov5s_v7_0-i800-e300.pt --source assets/selfie.jpg --imgsz 2048 --conf-thres 0.25 --iou-thres 0.45 --hide-labels --hide-conf

Maintainers

  • zhujian - Initial work - zjykzj

Thanks

Contributing

Anyone's participation is welcome! Open an issue or submit PRs.

Small note:

License

Apache License 2.0 Β© 2024 zjykzj