The neural network is in progress of learning to reconstruct frames of MLP episodes from smoothed 32x32 pixel images to 512x512 pixel ones. Since the network had no experience with either photos or unsmoothed images, inserting an unsmoothed photo of the plushie created this beautiful piece of art.
awesome, thanks. that looks amazing
I made a comparison video to demonstrate what it does when it is given a part of an episode.
https://youtu.be/_qtzGFYV–Q
Edited because: fixed link
Squints
My eyes aren’t working…