Creates a binary hdf5 training dataset file based on a training and validation text file listing matching videos and labels. Videos should have the same number of frames as the lines in the text file. Text file should contain tab-delimited columns containing the annotations from each labeler.
Manually edits labels in an existing dataset with full control on the majority of the parameters.
Script to add a video into the existing dataset. Defaults all the associated data (labels, mask, nlabelers) in as 0s.