Classic methods for inpainting instead of deep learning models

thygate / stable-diffusion-webui-depthmap-script

High Resolution Depth Maps for Stable Diffusion WebUI

MIT License

1.72k stars 159 forks source link

Classic methods for inpainting instead of deep learning models #339

Open darkcurrent opened 1 year ago

darkcurrent commented 1 year ago

When I look at the inpainting method, I see you are using various models. The results are perfect but it is really slow. May an option can be added in order to use openCV's own inpaint function like in this project. What do you say? Is this logical?

Cheers

semjon00 commented 1 year ago

Hello!

The first project I and @thygate knew about before, but it looks like it is hard to integrate into this repo. Would be cool if somebody could do that.

The second link though is something definitely worth checking out, thanks for pointing out!

Looking for a contributor for this one...

wandrzej commented 1 year ago

there was also One Shot 3D photography that supposedly did the inpainting in 1s , but the code never got released, only the depth estimation part. https://github.com/facebookresearch/one_shot_3d_photography/