takuseno / singan-nnabla Goto Github PK

Could you kindly add at least the following more examples:

• Generation from arbitrary scale
• Animation

Possibly add all images https://github.com/tamarott/SinGAN/tree/master/Input, depending on the total size and permission

• I am wondering about the unnecessary and redundant backward for the generator in this line of code:
  https://github.com/takuseno/singan-nnabla/blob/master/train.py#L230
• According to the paper, the already-trained sub-networks e.g., at lower scales than the current n-th scale, are not updated when we train the sub-network at n-th scale. This is implemented in train.py?
• In here, it is one-centered. Also, I am not sure why even the original author set axis=1 only.

Hi,

Thanks for the great implementation, it works so smoothly and quickly.

Do you have suggestions for how I would adapt it to perform image super-resolution?

Best wishes,

Mark

• Add references
• Add generation parts depending on types of applications, e.g, random sampling, super-resolution, animation, painting
• Training time w/ (single) GPU

Can anyone help, whole traceback is:
raceback (most recent call last):
File "main.py", line 16, in
Zs, reals, noise_amps = train(args)
File "D:\singan-nnabla-master\train.py", line 15, in train
reals = create_real_images(args)
File "D:\singan-nnabla-master\helper.py", line 68, in create_real_images
real_raw, args.min_size, args.max_size, args.scale_factor_init)
File "D:\singan-nnabla-master\helper.py", line 79, in calculate_scales
max_len = max([real_raw.shape[0], real_raw.shape[1]])
AttributeError: 'NoneType' object has no attribute 'shape'