Akashc6/aathraos-mobility

AathraOS Traffic Prediction Backend

This repository contains the backend API for the AathraOS smart city dashboard. It provides:

1. 1.Real-time Traffic Tracking: Headless CCTV YOLOv8 computer vision object counting (persons and vehicles).
2. 2.AI Peak Forecasting: A Random Forest machine learning model that analyzes historical metrics to predict weekly peak congestion hours.

Deployment Details

This project is configured to be deployed natively on Hugging Face Spaces as a Docker space.

The provided Dockerfile installs all necessary dependencies (including OpenCV headless libraries) and boots the FastAPI server via uvicorn using the start.sh configuration script. By default, it runs the fake_api.py data stream so it fits within free-tier hardware limits.

To run the actual YOLO CV tracking model on your own hardware:

1. 1.Clone the repository.
2. 2.Edit start.sh to run python3 -m uvicorn api:app --host 0.0.0.0 --port $PORT instead of fake_api.py.
3. 3.Provide a video or CCTV .mp4 stream directly to the /stream/start endpoint.