This model is based on landmark detection with ResNet18 backbone. Currently just work for 1 cat per image, and still wear glasses, because model just return position, and no confidence.
- Install the required packages
pip install requirements.txt
- Run real-time decorating with command
python decorate.py --source <0 for webcam, string for video path> --save_path <path to save the output video, default is ./output.mp4> --show_size <output video size, default is 640x480>
- Detecting multiple cats
- Improve the performance and accuracy
- Mode decorate options