apcuad / video-clipping

Data preparation for deep learning on video clip images

data-preparation-for-LAMWITTY

#####Setup brew install ffmpeg Check you've got Python already

#####Video clips

To create video clips, create directory called Footage with 3 directories inside named Y, N, and O and then run:

python supercut_creator.py [Movie file] [Annotation csv file]

Example: python supercut_creator.py p1.MOV p1.csv

#####Images from video clips

To create images, create directory called Images with 3 directories inside named Y, N, and O and then run:

python supercut_creator.py [Movie file] [Annotation csv file] true

Example: python supercut_creator.py p1.MOV p1.csv true

#####Shuffle videos

If you want to evenly separate participants between training and validation sets, you can use the shuffler.py script to randomly move clips from each participant to the appropriate folders. This is useful to check that we can train the model to predict reactions from the same people, but in the end we should have different participants in the training and validation sets.