Audit feature for QA/QC of labels

[nathanielrindlaub]

Develop this w/ Juli and Nick. Nick has suggested a couple ideas: one is a random selection of validated rodents and double check those, or “plant” a handful of rat images in the batch before a big label sprint, and see what % get flagged.

Most thorough would be to have some kind of “super confirm” / “retire” level of validation, and only high-permission level users would be able to do it, and Juli or some expert would be responsible for retiring ALL images labeled by volunteers.

If Juli doesn’t have capacity, maybe something along the lines of click a button and get a random sample?