Neural best-buddies: sparse cross-domain correspondence

Information

• Paper topic: Images
• Software type: Code
• Able to run a replicability test: True
• Replicability score: 5
• Software language: Python
• License: unspecified
• Build mechanism: Not applicable (python, Matlab..)
• Dependencies: CUDA / CuDNN / sklearn / numpy / torchvision / matplotlib
• Documentation score {0,1,2}: 1
• Reviewer: David Coeurjolly <david.coeurjolly@liris.cnrs.fr>
• Time spent for the test (build->first run, timeout at 100min): 20min

Source code information

• Code URL: https://github.com/kfiraberman/neural_best_buddies
• Git commit hash: 8121bb9b084026fc8dd097899f66e632c0f2fde3
• Software Heritage link:

Comments

sklearn dependency was not explicitly mentioned in the README. To make the code working on recent pytorch, I needed to update in line of code :


@@ -92,7 +92,7 @@ class vgg19(nn.Module):
             deconvolved_feature_forward = self.forward(level=src_level, start_level=dst_level, set_as_var = False)
             loss_perceptual = criterionPerceptual(deconvolved_feature_forward, src_layer)
             loss_perceptual.backward()
-            error = loss_perceptual.data[0]
+            error = loss_perceptual.item()
             self.update_last_losses(error)
             if (i % 3 == 0) and (print_errors == True):
                 print("error: ", error)

If you want to contribute with another review, please follow these instructions.

Please consider to cut/paste/edit the raw JSON data attached to this paper.