This is a pre-trained Yolo v5 for lips detection task. Just in case if someone will need it. I also provide labeled (by myself) data on which it has been trained (data_for_yolo folder). The original dataset is called AVLetters2.
Usage:
python detect.py --source gosling.jpg --weights lips.pt
Where gosling.jpg - input image (could be also a video, folder with images, your webcam and a lot of other inputs, check official YoloV5 repo for more information).
lips.pt - pretrained weights
The output will be in run/detect/expX folder.
Input:
Output:
You can use the weights for any purposes.