Lavanya14hs/Video-analysis-platform

# Video Analysis Platform A video incident analysis system with object detection, fight detection, accident detection, and crowd detection. ## Project structure - `api/` - FastAPI service exposing the video analysis endpoint - `backend/` - analysis logic and model utilities - `frontend/` - React UI for uploading and viewing results - `models/` - trained model files used by the analyzer - `logs/` - generated log files and snapshots - `temp/` - temporary upload storage ## Getting started ### Python backend 1. Create a Python virtual environment. python -m venv .venv .\.venv\Scripts\Activate.ps1 2. Install the required Python dependencies. pip install fastapi uvicorn opencv-python ultralytics numpy 3. Run the API server. uvicorn api.main:app --reload ### Frontend 1. Install frontend dependencies. cd frontend npm install 2. Start the React app. npm start ## Deployment ### Recommended: Railway 1. Create account at [railway.app](https://railway.app) 2. Install Railway CLI: `npm install -g @railway/cli` 3. Login: `railway login` 4. Initialize: `railway init` 5. Deploy: `railway up` The project includes `requirements.txt`, `Procfile`, `railway.json`, and a `Dockerfile` for deployment. The `Dockerfile` installs necessary system libraries for OpenCV in headless mode and is the recommended deployment path. ### Other Options See `DEPLOYMENT.md` for detailed instructions on Railway, Render, and Heroku deployment. ### Vercel Limitations Vercel is not suitable for this project because: - No Python runtime for ML models - Limited storage and compute resources - Cannot handle video processing workloads ## Notes - Keep your model weights in the `models/` folder, especially `yolov8n.pt`, `fight_detection.pt`, and `accident_detection.pt`. - Use `temp/` for uploads and clean it periodically. - Avoid committing large model files or sensitive keys to Git.