zjchust / MS-SSIM_L1_LOSS

Pytorch implementation of MS-SSIM L1 Loss function

MS-SSIM_L1_LOSS

Pytorch implementation of MS-SSIM L1 Loss function for image restoration.

How to use

import this .py file into your project.

from MS_SSIM_L1_loss import MS_SSIM_L1_LOSS
criterion = MS_SSIM_L1_LOSS()

# your pytorch tensor x, y with [B, C, H, W] dimension on cuda device 0
loss = criterion(x, y)

Please check demo.py for more details.

Requirements:

pytorch (only tested on pytorch 1.7.0 on windows 10 x64, other versions should work.)

Optional (only for testing this loss function)

numpy PIL

References:

[1] H. Zhao, O. Gallo, I. Frosio and J. Kautz, "Loss Functions for Image Restoration With Neural Networks," in IEEE Transactions on Computational Imaging, vol. 3, no. 1, pp. 47-57, March 2017

[2] https://github.com/NVlabs/PL4NN

[3] https://github.com/VainF/pytorch-msssim