We would like to improve network performance by adding new frames from unseen videos. Do you think the following steps are the way to go?

1. 'deeplabcut.add_new_videos()'
2. 'deeplabcut.extract_frames()' : change the config.yaml such that only new added videos are used for frame extraction i.e. comment out already seen videos.
3. 'deeplabcut.label_frames()'
4. train with best snapshot