Junlin Han

I think we simply ignored this " ASM seems to be designed for mixing a type of degradation and an image" and didn't consider their mixed effects from ASM.

Hello, remybonnav, Yes, you can definitely test a single model only. But this requires some changes in code (either in DCL model or putting DCL checkpoints/models to the CUT code),...

> Hello remybonnav, CUT model can support one direction translation (test A or B only). The architecture (ResNet-based Generator) of CUT and DCLGAN are identical. So if you want to...

Hello! Thanks for your questions. The optimizer_F uses default (pytorch) settings, I am not sure whether using the same hyperparameter as other optimizers will improve the results. You might have...

Hello @Sir-Zeeshan , Many thanks for your nice words! I'm currently quite busy with CVPR rebuttal and PhD application. I will provide a simple solution instead of writing a script,...

Hello Khawar, The ImageNet label is derived from SIMCLR. Please have a look at this link https://github.com/google-research/simclr

Hi, thanks for your nice words about our work. (1) The clean CityScape training set contains 2975 images, but the provided masks have different counts (rain streaks: 1620; raindrops: 3500;...

Hello khawar, Simply calculate (1 - top1 rate) would give the error rate. Note CIFAR-10 results can be unstable (that is, results can be a little bit different between multiple...