Inference example about pytorch-enhance HOT 4 CLOSED

Yeah it should actually be x = x.permute(1, 2, 0). I think the valid inputs of transpose may have changed over time. I'll make a fix for this and for checking that your input dims are correct as well.

from pytorch-enhance.

Fixed in #7 and #8. Removed the permute/transpose op since this breaks for batch of images. I was mostly using this to just quickly upres a single image and format it to be easily plotted but I think this is a better compromise.

from pytorch-enhance.

@robmarkcole What is the shape of test_images[0]? PyTorch models require a batch dimension so I would think that's the issue. Try doing sr_img = model.enhance(test_images[0].unsqueeze(0)).

from pytorch-enhance.

test_images[0].shape

torch.Size([1, 200, 200])

What is required is torch.Size([1, 1, 200, 200] so I try:

sr_img = model.enhance(torch.unsqueeze(test_images[0], 0))

getting

---> 80         x = x.transpose((1, 2, 0))
     81         return x

TypeError: transpose() received an invalid combination of arguments - got (tuple), but expected one of:
 * (name dim0, name dim1)
 * (int dim0, int dim1)
 * 

Might be a bug here? Pretty sure the transpose is not required

In this example they are using input = img_to_tensor(y).view(1, -1, y.size[1], y.size[0]) which achieves the same result..?

from pytorch-enhance.