The 'download.py' script can be used for the formal format of the CSV file, it downloads all videos for each playlist URL of podcast channels.
python download.py --videoList podcasts_list.csv
All stages (scenes detection, face detection, face tracking, face clips cropping, active speaker detection)
The 'ASD.py' script can be used to do the remaining stages
python ASD.py
The output files structure is illustrated below:
└── videos
├── podcast_# (folder for each podcast)
│ ├── videoIDd.mp4 (all videos in the podcast)
│ ├── metadata (folder to store metadata for each video)
│ ├── videoIDd.josn (josn file for all videos)
│
└── output
├── podcast_# (folder for each podcast)
│ ├── videoIDd (folder for each video on the podcast)
├── pycrop (The talking head videos and audios)
│ ├── 000000.avi
│ ├── 000000.wav
│ ├── 000001.avi
│ ├── 000001.wav
│ └── ...