Inspiration

We have all lined up the perfect shot only to have our pictures ruined by an innocent passerby. Our team decided that our beautiful pictures should not be compromised by people in the frame.

What it does

The application uses a recent deep-learning advancement called stable-diffusion to recognize and remove people from pictures, naturally filling in the places in the image where the people were.

How we built it

Python/Streamlit + Resnet-50 + StableDiffusion on GCP

Challenges we ran into

  • Python Streamlit integration with GCP database
  • Storing data using GCS
  • Google Cloud Engine Firewall issues to run our Streamlit app
  • Domain name rerouting
  • Streamlit stateful user control flow logic

Accomplishments that we're proud of

  • GCP Usage

What we learned

  • Streamlit
  • Deep learning
  • Cloud Computing

What's next for Empty World

  • Selective removal
  • Object removal

Built With

Share this project:

Updates