deblur-gan

deblur-gan copied to clipboard

Hello, can you please post both discriminator and generator weights. Thank you so much!

allavichik

hala3

In official paper and pytorch code, instance normalization is proposed after conv while batch normalization is used in this keras code. Is there any influence on the final result? Have...

zengqi0730

Something worng with testing my own image

4

Thanks for sharing your code! I got something wrong when test my own image. First, I changed the upsampling method in model.py to the older type to fix you gentrator.h5...

lx492008

Train issue:No module named keras.backend

1

`yff@Yff:~/PycharmProjects/Tf_DeblurGAN$ python scripts/train.py --n_images=512 --batch_size=16 --log_dir /path/to/log/dir Traceback (most recent call last): File "scripts/train.py", line 8, in from deblurgan.losses import wasserstein_loss, perceptual_loss File "/home/yff/PycharmProjects/Tf_DeblurGAN/deblurgan/losses.py", line 1, in import keras.backend as...

yff0815

how I add gradient_penalty_loss to the total loss? After many epochs, my d_on_g_losses is still relatively large, causing the deblur perfomance to drop. Could you tell how to implement the...

yl-precious

ModuleNotFoundError: No module named 'deblurgan'

5

When calling 'scripts/deblur_image.py', I receive the below:- ``` Traceback (most recent call last): File "/content/deblur-gan/scripts/deblur_image.py", line 5, in from deblurgan.model import generator_model ModuleNotFoundError: No module named 'deblurgan' ```

w3aryb0arpig

What are the training parameters used by the author while initially training the model?

shivam2296

How to choose 'epoch_num' in train.py to get a good result?

1

Thanks for sharing your code! My performance is not good, and the loss after each epoch is:  I want to know why my performance is not good. Is it...

ruixv

Can I train the model with multi GPUs?

2

Hi, I have two Nvidia cards, but when training the model, I see only one card is in use, although both cards' GPU memory is occupied. ```` root@bogon:~/DeblurGAN# nvidia-smi Wed...

wiwengweng