jasonwei20 / campus-gan Goto Github PK

Hi,
I took DCGAN-Tensorflow and trained on your dataset eleven campuses with a minor change that I kept the images in 128 size only and not resizing to 64.
python main.py --dataset eleven_campuses --input_height=128 --output_height=128 --train --epoch=270

My training ended with:
[ 4050 Epoch:[269/270] [ 14/ 15] time: 38037.0606, d_loss: 0.00000905, g_loss: 12.33740234
So I'm assuming training and checkpoint is fine.

I then generated samples using:
python main.py --dataset eleven_campuses --input_height=128 --output_height=128 --out_name="20210119.132128 - data - campus - x128.z100.uniform_signed.y128.b64" --visualize

But I'm getting just random pixels in each of the 100 output samples (which is 1024 image of 8x8 samples of 128 each)