search

Po-Hsun-Su / pytorch-ssim

pytorch structural similarity (SSIM) loss
Other
1.87k stars 364 forks source link

Multi-Scale Structural Similarity #9

Open jorge-pessoa opened 6 years ago

jorge-pessoa commented 6 years ago

Hello,

This SSIM implementation is a great addiction to the differentiable functions of PyTorch. However it is lacking the Multi-Scale Structural Similarity, a more commonly used metric in image assessment. I think it would be beneficial and relatively simple to implement and add this metric to this project.

I wrote a basic working MS-SSIM implementation based on SSIM from this project and using the Tensorflow implementation as reference. Would you consider adding the MS-SSIM metric if a pull request was sent?

Thank you,

chaos5958 commented 6 years ago

Dear jorge-pessoa,

I also need MS-SSIM, how is your progress? Could you share your version of MS-SSIM?

In addtion, I wonder whether there is GPU version of MS-SSIM regardless of using pytorch. (BUT SHOULD RUN ON PYTHON)

Thanks, Hyunho

jorge-pessoa commented 6 years ago

Hello,

I'm currently using the mentioned MS-SSIM implementation on training with success. The results obtained are equal to this implementation https://github.com/tensorflow/models/blob/master/research/compression/image_encoder/msssim.py within +- 4 decimal places so i suppose it is mostly correct.

I forked this repository and included the MS-SSIM implementation which you can find at https://github.com/jorge-pessoa/pytorch-msssim. I hope it is of some use to you.

Thanks, Jorge