What Image(s) Are Used for "Training" Computer Vision?

A post was split to a new topic: Use computer vision to annotate observations?