Redo json --> numpy --> bounding box outputs

[brialorelle]

• Code review on json --> numpy array code together; sanity check outputs with video frames
• Carefully strip videonames of extensions so easy for merging with other data (i.e. get rid of .AVI or .mp4, or .AVI.mp4 which does occur)
• Decide how to include bounding boxes of size zero in final outputs (and whether doable to include empty lines when no detedctions)
• Resave OP jsons outputs as numpy arrays with ALL people (not just first three in each frame)
• Backup all numpy arrays outside of sherlock (langcog server)
• Host large csvs from bounding box outputs on OSF or other permanent repo, and backup these output csvs on langcog server as well

[kachergis]

Any sticky issues here? What can we do to help?

[ketan0]

Hey—no sticky issues here. (Sorry for no commits over past couple days, have been busy with other work) I will be sure to backup the OP outputs to Langcog once they finish running (gonna start running Alice and Sam videos today.) I’m not as familiar with the bounding box code, so I have not approached those todo items.

[kachergis]

No worries on the bounding boxes -- I've got that! Mostly curious if you're happy and confident in json-->numpy output

[ketan0]

Ah awesome, thanks for doing that. I think that I am satisfied with the new condensed output format, as long as it's easy for you guys to manipulate in R (I find it to be so in Python.) Is it easy to load these dataframes with R's available packages (dplyr)?