pinto0309 / peppa_pig_face_landmark

A simple face detect and alignment method, which is easy and stable.

Peppa_Pig_Face_Engine

introduction

It is a simple demo including face detection and face aligment, and some optimizations were made to make the result better.

click the gif to see the video:

and with face mask:

requirment

• PyTorch
• onnxruntime
• opencv

model

1 face detector

yolov5-face

2 landmark detector

Simple keypoints detector.

WFLW	NME	Flops(G)	Params(M)	Pose	Exp.	Ill.	Mu.	Occ.	Blur	pretrained
Student@128	4.80	0.35	3.25	8.53	5.00	4.61	4.81	5.80	5.36	skps
Teacher@128	4.17	1.38	11.53	7.14	4.32	4.01	4.03	4.98	4.68	skps
Student@256	4.35	1.39	3.25	7.53	4.52	4.16	4.21	5.34	4.93	skps
Teacher@256	3.95	5.53	11.53	7.00	4.00	3.81	3.78	4.85	4.54	skps

I will release new model when there is better one.

By default student@256 is used in this project

install

git clone https://github.com/610265158/Peppa_Pig_Face_Landmark
python setup.py install

useage

# by code:

from Skps import FaceAna
facer = FaceAna()

result= facer.run(image)

## detect images, tracing is not used, add
##  facer.reset()

More details refer to demo.py

run python demo.py --cam_id 0 use a camera
or python demo.py --video test.mp4 detect for a video
or python demo.py --img_dir ./test detect for images dir no track

How to train

The codes are placed in folder TRAIN/face_landmark

Refer to TRAIN/face_landmark/README.md to train the model.