Welp. See all of you in church next Sunday

You trained 1 eye minions, 2 eyes minions, minions with glasses, or glasses on head, etc. without, I guess, describing enough differences. Or we miss trigger words. Because, as it is, most generations... well it's a yellow guy. But far from a minion.

It'd have maybe been better to focus your dataset on 1 kind of minion (same clothes, glasses always at the same place, etc.). To help the model understand the minion concept. Or maybe describe more the minions' images you have, with more detailed captions.

Anyway ;) Thanks for all your trainings.