Finish and polish the video-to-slots pipeline

[jasper-tms]

54ae4ae was a start but there are still steps to do. Files are going to
https://github.com/htem/gt_pickup_annotation/tree/main/video_to_slots

[jasper-tms]

Right now we extract every frame from the video, and the plan is to then select the best frames (the ones where the slots are closest to center) and throw out the rest. This is crazy wasteful, since saving every frame as a .png takes an enormous amount of storage space. Instead, can we extract every nth frame for some n as large as we can get away with? Instructions for pulling out only a subset of the frames from a video are at https://superuser.com/questions/135117/how-to-extract-one-frame-of-a-video-every-n-seconds-to-an-image but I haven't tried this myself yet

• Completed

[kuanawanda]

if i remember correctly we realized that this is not as trivial as it sounds because the speed of the tape is not strictly constant. So there needs to be some careful checks in here that probably involve barcode reading + manual checking of problem spots?

[Wangchu-Xiang]

I am extracting some frames from pickup video and I think I can work on the wiki page to add more guidance on extracting frames and crop slots.