Our inspiration for the current project came from the need to address a key problem: existing tools struggle with noisy photos and videos, where pixel differences often trigger false alarms. We aim to build a pipeline that flags only meaningful changes and can run efficiently on regular consumer hardware.
Our goal is to implement an efficient, lightweight, and explainable ML pipeline that accepts two frames, aligns them, computes an SSIM-based change map, and returns an overlay.
Built With
- fastapi
- numpy
- opencv
- python
- react
- scikit-image
- typescript
- typescript-backend:-fastapi
- uvicorn
- uvicorn-vision/ml:-opencv
Log in or sign up for Devpost to join the conversation.