search

up42 / image-similarity-measures

:chart_with_upwards_trend: Implementation of eight evaluation metrics to access the similarity between two images. The eight metrics are as follows: RMSE, PSNR, SSIM, ISSM, FSIM, SRE, SAM, and UIQ.
MIT License
549 stars 68 forks source link

Numpy optimizations #52

Closed ajinkya-kulkarni closed 1 year ago

nekhtiari commented 1 year ago

This could potentially help us with this issue. But having some metrics for this would be good to see if it really does the optimisation!

ajinkya-kulkarni commented 1 year ago

I could try comparing two images with and without these changes. I'll update the results here when I get time. Thanks for the suggestion!

seedlit commented 1 year ago

@ajinkya-kulkarni I added some tests. I tested your implementation (uiq) it is around 3x faster. Nice! Can you pull changes from master?

ajinkya-kulkarni commented 1 year ago

Thanks for the tests! I have synced my fork with the master branch.

seedlit commented 1 year ago

@ajinkya-kulkarni could you resolve the merge conflicts?