Richard Zhang

I updated the pip, so hopefully it should work now. Thanks!

Sorry, I probably should have implemented it to be like this to begin with, and then the linear layers wouldn't have to be clamped during training. But the linear layers...

We have some experiments in suppmat (Figure 12), actually observing worse correlation with resnet. Interestingly, this is consistent with some results in neuroscience on correlating IT responses with deep network...

Thanks for trying to reimplement! The averaging across feature channels in the last layer seems potentially problematic. See Table 6 for the architecture here: https://arxiv.org/pdf/1611.09842.pdf . Thanks!

Thanks for cleaning up the code! Could we get the original font back?

Please see [this script](https://github.com/richzhang/PerceptualSimilarity/blob/master/compute_dists_pair.py) for LPIPS distance on pairs of images in a directory. I added a description of the procedure here: https://github.com/junyanz/BicycleGAN#metrics.

Thanks for the question. The discrepancy is caused by the version of LPIPS. BicycleGAN used v0.0. This version turned out to have a preprocessing bug. It is kept as a...

Great. I was trying the ground truth images when responding to this. Thanks

Hi, are you using the caffe weights converted into PyTorch?

Technology based on this is also now in Photoshop Elements 2020