You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It's container/story for all activities related to adding video support. The following things should be considered:
Splitting videos into frames
Extracting metadata only from main - video file (ignoring frames)
Keeping information about parent file for each frame
Main goal for splitting frames is to provide image format for Metadata Enrichment (Deep Learning). User should have functionality of detecting catagories considering whole video (frame by frame).
Potential performance issues should be considered - loading whole video to DataFrame and then splitting it causes problems with processing long/big videos. We can add repartition after splitting but this will cause shuffling which is definitely not good. Probably it will be reasonable to provide user option (in config) to use one method or another. Much better solution is custom Data Source which will be splitting video on read. However it's not so easy and still images should be considered here also as user could point to directory with mixed data: jpg, avi, mp4, etc.
It's container/story for all activities related to adding video support. The following things should be considered: