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:

• PyTorch
• onnxruntime
• opencv
• python 3.7
• easydict

• 1 face detector

  yolov5-face

• 2 landmark detector

  simple face landmark detector

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

  the model is trained with WFLW data. For student mobilenetv3-large was used as backbone, for teacher is efficientnetb5.

  model	Resolution	NME(test set)	model size (int8 weights)	Pretrained
  Student	128x128	4.95	1.9M	model128
  Teacher	128x128	4.64	6.9M	model128
  Student	256x256	4.65	1.9M	model256
  Teacher	256x256	4.47	6.9M	model256

UPDATE Better model

I will release new model when there is better one. 7.5K trainning data is not enough for a very good model. Please label more data if needed.

1. pretrained models are in ./pretrained, for easy to use ,we convert them to mnn
2. 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
   or python demo.py --video test.mp4 --mask True if u want a face mask

# by code:
from lib.core.api.facer import FaceAna
facer = FaceAna()
boxes, landmarks, _ = facer.run(image)