a gan attempt modified from the tensorflow example that generated digits, still don't really know what i'm doing at all. this network was traumatised by 500 specimens of imagenet's finest cuisine

pretty much expected it to not work but i wanted to see what would happen. it is more clear to me now why you can't just train it to generate an image of anything, thinking of the types of gans i've seen already

the gangogh project i saw was pretty cool, and i would like to understand how that works. deep dream also inspired me when i saw it years ago, there's still so much to learn

the animation shows what the network generates from the same random seed as it learns over time. my favourite one is the second more wavy wafer, i enjoy looking at it. i think the third and fourth that developed more distinct patterns after that were when i added more layers, which would make sense