Skip to content

Latest commit

 

History

History
27 lines (16 loc) · 1.13 KB

README.md

File metadata and controls

27 lines (16 loc) · 1.13 KB
Raw

Partial Convolution Web Demo

To try it out, clone, put this file in /backend/, and run:

pip install -r requirements.txt

npm i

npm start

This will concurrently run the React frontend and Flask backend.

The frontend allows you to upload an image and erase parts of it with the mouse. Then when you click "infill", the image and mask are sent to a PartialConvUNet in the backend which infills your image using AI.

References

This is an unofficial web demo of a paper, Image Inpainting for Irregular Holes Using Partial Convolutions [Liu+, arXiv2018].

These existing open-source implementations were referenced extensively, and we are using pretrained weights from one of them:

https://github.com/NVIDIA/partialconv

https://github.com/naoto0804/pytorch-inpainting-with-partial-conv

https://github.com/tanimutomo/partialconv