Could you release the code for your adapted evaluation metrics (FID)?

NVIDIA / vid2vid

Pytorch implementation of our method for high-resolution (e.g. 2048x1024) photorealistic video-to-video translation.

Other

8.58k stars 1.2k forks source link

Open donghaoye opened 5 years ago

donghaoye commented 5 years ago

help

donghaoye commented 5 years ago

We recommend using a minimum sample size of 10,000 to calculate the FID otherwise the true FID of the generator is underestimated.

But, in vid2vid, there are only 500 videos for testing. How did you calculate the FID score?

donghaoye commented 5 years ago

In addtion, "For each video, we extract a spatio-temporal feature map with this CNN. " It means that 500 samples for FID?

Bowenwu1 commented 5 years ago

linkAmy commented 5 years ago

MannyKayy commented 5 years ago

UmarSpa commented 2 years ago