Goulustis

I was testing the example images pair: ./imgs/ex_p0.png ./imgs/ex_p1.png I found the output of lpips is non-deterministic with dropout turned off. My further investigation reveals that the linear tuned weights...

"indices" and "batch["image_idx"]" can be on different devices. The assignment between them, though will allow the model to converge correctly, have unpredictable behaviors. Namely, indices[:, 0] - batch["image_idx"][c].cpu() != 0....

Hi, I was wondering if you could show some video samples generated using your method? I was digging to find some supplementary materials in your paper and found none.... Thanks...