search

avalonstrel / GatedConvolution_pytorch

A modified reimplemented in pytorch of inpainting model in Free-Form Image Inpainting with Gated Convolution [http://jiahuiyu.com/deepfill2/]
Other
434 stars 77 forks source link

about the package version #34

Closed wdrink closed 4 years ago

wdrink commented 4 years ago

Hi, thanks for your work in implementing the code with pytorch. I encountered an error that the netD will crash if using 64*64 as size of input img and mask:

RuntimeError: Calculated padded input size per channel: (3 x 3). Kernel size: (4 x 4). Kernel size can't be greater than actual input size

The same code works well when I set the size to 128128 or 256256.

Francoforever commented 3 years ago

请问你解决了嘛？遇到同样问题 求教 多谢~~

I solved. To modify these 2 lines to 256*256. https://github.com/avalonstrel/GatedConvolution_pytorch/blob/22455dae24c42b58ac3fe4803c35487fe56d56dd/test_images.py#L33-L34 FYI~