Inspiration

Nvidia research

What it does

Upload a (square-ish) photo. Click an object (wait 10 seconds). Watch it disappear!

It combines https://www.nvidia.com/research/inpainting/

with https://github.com/matterport/Mask_RCNN

We combine instance segmentation with inpainting to automatically detect and remove objects from photos

How I built it

GCP tensorflow flask js scipy pytorch

Challenges I ran into

  • python config
  • cuda setup
  • tensorflow setup
  • gpu setup
  • pytorch

Accomplishments that I'm proud of

it works! (please don't overuse the URL, I only have a single gpu)

What I learned

I read a few papers and implementations

What's next for SceneClean

faster feed forwards, cleaner interface, caching

Built With

Share this project:
×

Updates