pytorch-msssim

Differentiable Multi-Scale Structural Similarity (SSIM) index

This small utiliy provides a differentiable MS-SSIM implementation for PyTorch based on Po Hsun Su's implementation of SSIM @ https://github.com/Po-Hsun-Su/pytorch-ssim. At the moment only a direct method is supported.

Installation

Master branch supports PyTorch 0.3.1 or lower. For PyTorch 0.4 support use the dev branch.

Go to the repo directory.
Run python setup.py install

Clone this repo.
Copy "pytorch_msssim" folder in your project.

Example

The provided images for calculation must be Variables.

basic usage

import pytorch_ssim
import torch
from torch.autograd import Variable

m = pytorch_msssim.MSSSIM()

img1 = Variable(torch.rand(1, 1, 256, 256))
img2 = Variable(torch.rand(1, 1, 256, 256))


if torch.cuda.is_available():
    img1 = img1.cuda()
    img2 = img2.cuda()

print(pytorch_msssim.msssim(img1, img2))
print(m(img1, img2))

Reference