BYchao100 / Deep-Lossy-Plus-Residual-Coding

Deep Lossy Plus Residual Coding for Lossless and Near-lossless Image Compression (TPAMI'24, Journal Extension of our CVPR'21 work)
14 stars 0 forks source link

PSNR and max_diff_nll for different τ values #4

Open 1564125986 opened 4 months ago

1564125986 commented 4 months ago

Hello, I would like to ask why psnr will become smaller with the decrease of τ value, while max_diff_nll value remains at about 50, which does not become smaller. image

BYchao100 commented 4 months ago

Hi there, I think your code might be running incorrectly. When $\tau=0$, there should be no loss, PSNR should be infinite, and max_diff should be $0$. Something seems to be off here. Could you check if your experimental environment matches the recommended one, such as the PyTorch and CUDA versions?