search

fnzhan / MoNCE

[CVPR 2022] Modulated Contrast for Versatile Image Synthesis
MIT License
35 stars 5 forks source link

questions about swd #4

Open haoren55555 opened 1 year ago

haoren55555 commented 1 year ago

Sorry to bother, I saw you used SWD (sliced wasserstein distance) in the paper, and I also want to use this score as a metric for some tests. Could you offer your code to calculate it? thanks!

fnzhan commented 1 year ago

please refer to Utils/SWD/cal_sliced_wasserstein.py.

haoren55555 commented 1 year ago

thanks!

haoren55555 commented 1 year ago

well, i tested the code, and it seemed that the code somehow needs a "sliced_wasserstein.API", and could you tell me where to get it?

fnzhan commented 1 year ago

please refer to Utils/SWD/sliced_wasserstein.py

haoren55555 commented 1 year ago

I tested the code and it worked perfectly well. The last question I want to ask is that there are six scores "SWDx1e3_256 SWDx1e3_128 SWDx1e3_64 SWDx1e3_32 SWDx1e3_16 SWDx1e3_avg" there, and it seemed to use the last one, right? also the two image folders I should use is the fake_B(generated_zebra, for example) and real_B(real_zebra, for example), right?

fnzhan commented 1 year ago

yes, as you described.