devcam

Easy local camera feeds for development and testing. Expose a webcam as HTTP MJPEG or RTSP, with an optional built-in YOLO human detection overlay — all from one command.

Install

pip install devcam

To enable human detection (YOLOv8), install with the vision extra:

# quote the brackets in zsh
pip install 'devcam[vision]'

Quick Start

devcam http            # web UI + MJPEG stream on http://localhost:1080
devcam rtsp            # RTSP stream on rtsp://localhost:8554/cam

Open http://localhost:1080/ to view the live stream, switch cameras, change resolution/FPS, mirror the view, and toggle human detection. Detection is lazy-loaded: the YOLO model only loads when you first enable it.

Usage

devcam {http,rtsp} [--port PORT] [--camera INDEX] [--resolution WxH] [--fps FPS] [--backend {auto,cv2,zed}]

HTTP Endpoints

Human Detection Endpoints

Requires devcam[vision]. Without it, streaming works normally and detection endpoints return available: false.

Example:

curl -X POST http://localhost:1080/api/detection/config \
  -H 'Content-Type: application/json' \
  -d '{"enabled":true,"confidence":0.55,"interval_ms":50}'

Detection response:

{
  "human_detected": true,
  "confidence": 0.87,
  "count": 1,
  "latency_ms": 42.1,
  "detections": [
    {"confidence": 0.87, "bbox": [420.0, 160.0, 780.0, 940.0]}
  ]
}

Camera Backends

Jetson / ZED X Notes

• If ZED cameras show as "NOT AVAILABLE": sudo systemctl restart zed_x_daemon
• Resolution opens at the camera's native mode; --resolution downscales in software
• --fps 5 or --fps 15 recommended for low-bandwidth use cases

RTSP Setup

RTSP requires ffmpeg and mediamtx. One-time setup:

# macOS
brew install ffmpeg && bash setup_mediamtx.sh

# Linux
sudo apt install ffmpeg && bash setup_mediamtx.sh

Then each time:

./mediamtx &             # media server in the background
devcam rtsp

Development

git clone https://github.com/GitAashishG/devcam.git
cd devcam
pip install -e '.[dev,vision]'
python -m pytest tests/ -v

Releases happen automatically when the version in pyproject.toml is bumped and merged to main.

macOS Notes

• Camera permission: System Settings → Privacy & Security → Camera → Terminal
• All ports are unprivileged (≥ 1024) — no sudo needed

License

MIT