search

richzhang / PerceptualSimilarity

LPIPS metric. pip install lpips
https://richzhang.github.io/PerceptualSimilarity
BSD 2-Clause "Simplified" License
3.58k stars 493 forks source link

How I can derive ResNet model weight from the pre-trained LPIPS weights #133

Open dinusha94 opened 3 months ago

dinusha94 commented 3 months ago

Hi, in the Nvidia Stylegan2-ada they have mentioned that they use vgg16 weights which are derived from the pre-trained LPIPS weights,

"vgg16_zhang_perceptual.pkl" is further derived from the pre-trained LPIPS
weights by Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman

link : https://nvlabs-fi-cdn.nvidia.com/stylegan2-ada/pretrained/metrics/NOTICE.txt

I would like to do the same for other networks also, How can I do that

Thanks in advance Dinusha

unistdJRZ commented 2 months ago

Perceptual Loss has specified model structure and trained by specified dataset，if u wanna change into ResNet u should doing this procedure again