Zhaoyi-Yan

PatchSoft means `shift_sz` not only be 1\*1, but also can be 3\*3, etc. It is equivalent to contextual attention. `res_xx` means that I do not directly concat the shifted feature...

Hi, in fact it is a trick to define the masked region in feature space. Always, the simplest way is directly resizing the mask. However, in fact the border of...

For the first question, #6 may help you. For 2nd question, all ground-truth images should be placed in one folder. That's all ! You should set `mask_type=random` in training. When...

As the mask of line is quite thin. It is not quite necessary for gan training. As shift operation shift pixels from known region to missing region, it is not...

try your `display_id=0` to disable the visualization during training. Maybe it is the incompatibility of visdom?

try `display_port=9001`, if it still occurs an error, post the error message here.

Are you sure when you change https://github.com/Zhaoyi-Yan/Shift-Net_pytorch/blob/master/options/train_options.py#L12 to something like `9001`, it still sends you address 8097 even if your running script does not contains `display_port` item?

Maybe you are running the code on a server which contains no screens, you may try https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix , and ask for help there.

Hi, our original code is base on `Torch7`, https://github.com/Zhaoyi-Yan/Shift-Net . We haven't got that performance in pytorch. But we promise, there is no much gap between these two versions.

Hi, I recommend you try our implemetation based on pytorch (Not this Torch version), which is much faster. I will take less than 1 day when batchsize=1 for training 30...