Skip to content

Latest commit

 

History

History
53 lines (29 loc) · 1.32 KB

README.md

File metadata and controls

53 lines (29 loc) · 1.32 KB

chainer SSIM

Differentiable structural similarity (SSIM) index.

This repository is based on Po-Hsun-Su/pytorch-ssim.

Code of generation Einstein from random image is here

Note that this code have not tested on cpu.

Requirements

  • chainer (6.0.0)
  • cupy (6.0.0)

How to use

Please see baseline_demo.py

python baseline_demo.py

Difference between baseline_demo.py and structural_similarity2d_loss_demo.py?

It's just whether code use 1d decomposition for gaussian filter.

I recommend to use structural_similarity2d_loss_demo.py.
Since it's fast 😄

3D result

Please see structural_similarity3d_loss_demo.py

python structural_similarity3d_loss_demo.py
Target image Initial image Reconstructed image

Reference