weizmannvision / dropthegan Goto Github PK

The image_utils in integration_test.ipynb is missing, can you please add it?

Thanks for your great work about single image generation.
Could you release the datasets of SIGD16? Thanks.

Hi, I just try to use the application of "conditional_inpainting" and it looks like the mask was not properly readed:

$ python main.py conditional_inpainting --masked_image_path="data/conditional_inpainting/balloons_in_blue.png" --mask_path="data/conditional_inpainting/mask.png" --output_path="out/conditional_inpainting/balloons.png"

Traceback (most recent call last):
File "/home/machine/lcg/projects/DropTheGAN-main/main.py", line 196, in
fire.Fire()
File "/home/machine/.conda/envs/dropthegan/lib/python3.10/site-packages/fire/core.py", line 141, in Fire
component_trace = _Fire(component, args, parsed_flag_args, context, name)
File "/home/machine/.conda/envs/dropthegan/lib/python3.10/site-packages/fire/core.py", line 466, in _Fire
component, remaining_args = _CallAndUpdateTrace(
File "/home/machine/.conda/envs/dropthegan/lib/python3.10/site-packages/fire/core.py", line 681, in _CallAndUpdateTrace
component = fn(*varargs, **kwargs)
File "/home/machine/lcg/projects/DropTheGAN-main/main.py", line 151, in conditional_inpainting
output_image = applications.conditional_inpainting(masked_image,
File "/home/machine/lcg/projects/DropTheGAN-main/applications.py", line 61, in conditional_inpainting
return gpnn.gpnn(pyramid,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 36, in gpnn
generated = pnn(generated,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 91, in pnn
fold.unfold2d(mask, patch_size))
File "/home/machine/lcg/projects/DropTheGAN-main/fold.py", line 31, in unfold2d
raise ValueError('expects a 4D tensor as input')
ValueError: expects a 4D tensor as input

I tried to modify the codes of "mask_read()" in utils.py to keep the dimension of the mask and got another error:

Traceback (most recent call last):
File "/home/machine/lcg/projects/DropTheGAN-main/main.py", line 196, in
fire.Fire()
File "/home/machine/.conda/envs/dropthegan/lib/python3.10/site-packages/fire/core.py", line 141, in Fire
component_trace = _Fire(component, args, parsed_flag_args, context, name)
File "/home/machine/.conda/envs/dropthegan/lib/python3.10/site-packages/fire/core.py", line 466, in _Fire
component, remaining_args = _CallAndUpdateTrace(
File "/home/machine/.conda/envs/dropthegan/lib/python3.10/site-packages/fire/core.py", line 681, in _CallAndUpdateTrace
component = fn(*varargs, **kwargs)
File "/home/machine/lcg/projects/DropTheGAN-main/main.py", line 151, in conditional_inpainting
output_image = applications.conditional_inpainting(masked_image,
File "/home/machine/lcg/projects/DropTheGAN-main/applications.py", line 61, in conditional_inpainting
return gpnn.gpnn(pyramid,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 36, in gpnn
generated = pnn(generated,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 108, in pnn
_, indices = find_normalized_nearest_neighbors(query_patches_column,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 239, in find_normalized_nearest_neighbors
return find_weighted_nearest_neighbors(queries, keys, normalizer,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 200, in find_weighted_nearest_neighbors
tile_height = _find_tile_height(num_queries, num_keys,
File "/home/machine/lcg/projects/DropTheGAN-main/gpnn.py", line 163, in _find_tile_height
return min(height, (max_tile_size + row_size - 1) // row_size)
ZeroDivisionError: integer division or modulo by zero

Could you please give me some help to solve this error :)
Thanks

Hi,

Thanks for the great work! I am playing with your code and it worked remarkably well. However, I do notice some artifacts for the balloon example: when I use the code to do either random generation or retargeting, it always shows this "glowing" effect around the balloons like this. This artifact is not shown in the original image or in your paper. I am wondering what might be wrong. Thanks for your help in advance!