Gack

Privacy-preserving home monitoring using pose detection.

Features

• Real-time pose detection from RTSP streams
• Privacy-preserving (only stores pose data, not video frames)
• SQLite database storage for detection replay
• Web interface for viewing historical detections
• Configurable detection frequency and output options

Installation

1. Install dependencies:

pip install -e .

2. Create a .env file with your configuration:

UNIFI_RTSPS_URL=rtsp://your-camera-url
FPS=1
SHOW_ORIGINAL=False
SAVE_TO_DB=True
LOG_LEVEL=INFO

Usage

Running Pose Detection

Start the pose detection stream:

gack

This will:

• Connect to your RTSP stream
• Run pose detection at the specified FPS
• Save detection data to SQLite database (if enabled)
• Output processed video to outdata/output.mp4

Web Interface

Start the web interface to view historical detections:

gack-web

Then open http://localhost:8000 in your browser.

The web interface provides:

• Database statistics
• Browse latest detections
• Search detections by time range
• Visual replay of pose detections
• Interactive canvas showing skeletons and keypoints

Environment Variables

• UNIFI_RTSPS_URL: RTSP stream URL for your camera
• FPS: Detection frequency (default: 1)
• SHOW_ORIGINAL: Show original video background (default: False)
• SAVE_TO_DB: Save detections to database (default: True)
• LOG_LEVEL: Logging level (default: INFO)
• WEB_HOST: Web interface host (default: 0.0.0.0)
• WEB_PORT: Web interface port (default: 8000)

Database Schema

The SQLite database stores:

• Detection timestamps (ISO 8601 in UTC, e.g., 2024-01-30T12:34:56Z)
• Frame numbers and video timestamps
• Pose keypoints (17 points per person)
• Detection confidences
• Bounding boxes
• Keypoint confidence scores

API Endpoints

• GET /api/stats - Database statistics
• GET /api/detections/latest?limit=50 - Latest detections
• GET /api/detections/timerange?start_time=...&end_time=... - Detections by time range
• GET /api/detections/{id} - Specific detection by ID

Development

Run tests:

pytest tests/

Privacy

This system is designed for privacy preservation:

• Only pose keypoints are stored, not video frames
• No facial recognition or personal identification
• Detection data can be easily deleted from the database
• Local storage only (no cloud uploads)