search

Maclory / SPSR

Pytorch implementation of Structure-Preserving Super Resolution with Gradient Guidance (CVPR 2020 & TPAMI 2021)
448 stars 83 forks source link

Errors accur when using evaluation tool box #32

Closed HarrysVision closed 3 years ago

HarrysVision commented 3 years ago

Hi, when I using the evaluation tool box to test one of the benchmark datasets, an error accurred:

Traceback (most recent call last): File "evaluate_sr_results.py", line 99, in LPIPS=CalLPIPS(j,k) File "evaluate_sr_results.py", line 56, in CalLPIPS dist = model.forward(imageA,imageB).detach().squeeze().numpy() File "G:\dissertation\New Papers\SPSR\SPSR-master\metrics\LPIPS__init__.py", line 40, in forward return self.model.forward(target, pred) File "G:\dissertation\New Papers\SPSR\SPSR-master\metrics\LPIPS\dist_model.py", line 117, in forward return self.net.forward(in0, in1, retPerLayer=retPerLayer) File "G:\dissertation\New Papers\SPSR\SPSR-master\metrics\LPIPS\networks_basic.py", line 72, in forward diffs[kk] = (feats0[kk]-feats1[kk])**2 RuntimeError: The size of tensor a (170) must match the size of tensor b (169) at non-singleton dimension 2

I follow the instructions provided, and other datasets seems to work well( I tested on Set5, B100, etc.), why?

Maclory commented 3 years ago

Hi, the sizes of the SR image and the corresponding HR image may be different. In general, the SR and HR image sizes should be a multiple of 4 since the upscaling factor is 4.