This looks like the inverse of the usual "tell we what's in this video" design - Dextro seems to want you to ask it if something their system already knows about is in a video.
I think I'd rather just submit a video and have all the matches and their confidence / salience (not sure that these are the same thing).
I think I'd rather just submit a video and have all the matches and their confidence / salience (not sure that these are the same thing).