vaegan
vaegan copied to clipboard
On the original code the recon_depth is 9.
I feel that the image_feature should be h4 before activation function and batch normalization. Did you guys try to do it based on the paper writer's code?