Face_Lip_detection_tracking

립리딩 영상 face / lip ROI labelling 작업

1. Example

2. How to run

• Locate videos to be processed at ./sample (args.root_dir)
• Results will be saved at ./out (args.save_dir)
• If error occurs, file_name and error message will be written at error_list.txt (args.error_txt_path).

python main.py

3. Output directory

• Running main.py will detect face ROI and lip ROI at the same time.
• ./out directory tree is as follows.

├── lip_mp4
   ├── sample1.mp4
   ├── sample2.mp4
   └── sample3.mp4
             '
             '  
             '
├── lip_json
   ├── sample1.json
   ├── sample2.json
   └── sample3.json
             '
             '  
             '
├── face_mp4
   ├── sample1.mp4
   ├── sample2.mp4
   └── sample3.mp4
             '
             '  
             '
├── face_json
   ├── sample1.json
   ├── sample2.json
   └── sample3.json

4. Details

Face_Lip_detection_tracking_guide

5. lsh branch

• KLT tracker version (main_KLT.py, only for face tracking) added.
• Python logger used instead txt file.
• Input configs(*.yaml) for args variables setting.
• Exception handling regarding median flow tracker success flag added (main_FAIL.py).