Not meaning to sound too critical because it's cool but there's so little information here it's hard to interpret?

I'm kinda wondering about the nature of discrepancies between the inferred and actual images, and what the training process actually entailed.