Jack Yi

icon indicating a website link https://scholar.google.es/citations?user=XC6Z56AAAAAJ&hl=en icon indicating an email [email protected]

icon company Unknown icon location Vancouver, BC, Canada icon location Jack Yi started programming in 2007. His mission is to make the creation of visual contents easier.

Results 19 comments of Jack Yi

Scipy image IO deprecated

Thanks for the suggestion. I will update the codes ASAP.

When will the training step stops?

when the epoch_idx reachers args.epoch. If you want to stop it early, you may use smaller number for args.epoch.

segmentation fault

I have never saw that error before. It is probably due to lack of swap memory in your computer.

About the training loss

The result images shown in the main body of the paper are generated with WGAN. After the acceptance of the paper, I did some further experiments and found that the...

About the training loss

Instead of explicitly adding random noise, I used dropout instead.

convergence of training process

No, convergence means the image quality reaches satisfactory. Practically, you need to specify the args.epoch manually by observing the image quality.

Does not work for non jpg images

Thanks for the feedback, I will revise the codes soon. On 24 October 2017 at 14:10, kickbox wrote: > I found that the "model.py" code has been written to search...

where does the 'center_crop()' in the utils.py import from?

It is a left over. Sorry for causing the confussion. I will update the codes then.

Can you please upload the trained models for DualGAN?

I will update the codes and also upload the pretrained models ASAP. Thanks for the request.

Where is the ‘clip_trainable_vars‘ used?

Yes. I have updated the codes a few times and disabled wgan loss Sorry for causing the confusion. In the near future, I will provide both options (wgan and nsgan)...