Looks neat. But ya gotta watch your captions better. I noticed from opening just one, that you tag things that aren't even in the given image, and don't tag some things that are in the image.

For example; the image with (what looks like) some form of holographic smartwatch; you didn't tag that at all, and given that it's right in the middle of the image/character, that's likely to confuse the AI during training. But, you did tag his belt and such that wasn't in that image.