Model 1 - 1,000 images [POC_adapted_small.py] DCGAN Dense, upsampling x 2, conv2d, output conv2d sigmoid 4 discri
Model 1 - 80,000 images [POC_adapted_28_aws.py] DCGAN Dense, upsampling x 2, conv2d, output conv2d sigmoid 4 discri
Model 2 - 80,000 images [POC_increase_1_aws.py] DCGAN Dense, upsampling x 4, 2 maxpooling, conv2d, output conv2d sigmoid 4 discri
Maxpool 2x2 and stride 2 reduces the image exactly into half. goes in a pair with an upsampling! https://wiseodd.github.io/techblog/2016/07/18/convnet-maxpool-layer/
Model 3 - 80,000 images [POC_increase_2_aws.py] DCGAN Dense, upsampling x 4, 2 maxpooling, conv2d, output conv2d sigmoid 6 discri
Adv loss goes to 0 very quickly
Model 4 - 80,000 images [POC_250118.py] DCGAN Dense, upsampling x 2, conv2d, output conv2d tanh 4 discri