search

rgcda / haarpsi

The Haar wavelet-based perceptual similarity index (HaarPSI) is a similarity measure for images that aims to correctly assess the perceptual similarity between two images with respect to a human viewer.
MIT License
35 stars 5 forks source link

so computational #2

Open aligoglos opened 4 years ago

aligoglos commented 4 years ago

I'm interesting to this work but its too slow for real time applications. I think if you can re implement it in cudaand c++, it will be great.

zakajd commented 4 years ago

Hi I'm not this repo developer, but maintain another library with collection of metrics github.com/photosynthesis-team/piq Just added HaarPSI metric, feel free to use it. It's written on PyTorch framework, which allows GPU computations. You can also compute HaarPSI for a batch of images. Hope it will be useful :)

jowlo commented 4 years ago

Just out of interest, why would you want to use a PSI in a real time application?