ImageDecomposer is an e2e app that detects objects in an input image and outputs them as separate images. Backend is constructed using FastAPI, frontend - streamlit. The underlying mechanism involves Mask R-CNN implemented in detectron2 library.
Input an image
And the program outputs segmented instances as separate images
To run the app on your own device clone the repository, unzip fastapi/detectron2. Then in a machine with Docker run commands:
docker-compose build
docker-compose up
Visit http://localhost:8501.
In this project I use detectron2 library developed by Facebook AI Research team.