- Dataset : https://google.github.io/cartoonset
- batch_size : 64
- H, W = [256, 256]
- Discrminator_lr = 0.00025
- Generator_lr = 0.0002
- Weights init = glorot_uniform (to be changed to RandomNormal[mean=0, stddev=0.02])
- Discriminator training strategy : separate weight updates for real and fake batch
srihari-humbarwadi / cartoon_gan_tf2.0 Goto Github PK
View Code? Open in Web Editor NEWA DCGAN implementation in tensorflow 2.0, trained on cartoonset100k dataset