This project provides real-time crowd analysis using a drone video feed. The stack consists of:

• MediaMTX: Local RTMP ingest & WebRTC streaming server.
• CSRNet-pytorch: Python backend for AI inference and stream info.
• Frontend: React application for visualization.

• Docker & Docker Compose
• Node.js & npm
• Python 3.9+

Run the helper script for your OS:

Mac/Linux:

./scripts/dev_up.sh

Windows (PowerShell):

./scripts/dev_up.ps1

This will:

1. Start MediaMTX in Docker (Background).
2. Ask if you want to run Backend & Frontend (Foreground/New Windows).

./scripts/dev_up.sh

This starts MediaMTX (RTMP Server), Backend, and Frontend.

Verification: After running the script, verify that the media server is up:

lsof -i :1935

You should see mediamtx listening on port 1935.

1. Connect your DJI drone to the Same Wi-Fi as your computer.
2. Open the dashboard at http://<YOUR_LAN_IP>:3000.
3. Copy the RTMP Ingest URL (e.g., rtmp://192.168.1.16:1935/dji).
4. Paste it into the DJI App (RTMP Streaming).
5. Start Streaming.

Note on Backend Logs: You may see warnings like [WARN] RTMP server not reachable... if the media server is not up yet. This is normal. The backend will retry automatically.

• Frontend: Open http://localhost:3000 (or http://<LAN_IP>:3000 from another device).
• Direct Player: http://localhost:8889/dji

• 1935: RTMP Ingest (Input)
• 8889: WebRTC Playback (Output)
• 8000: Backend API
• 3000: Frontend

• No Video? Ensure DJI is sending H.264 video (not H.265).
• Connection Failed? Ensure Firewall allows ports 1935 and 8889.