Frigate

Frigate is an open-source, self-hosted network video recorder designed for local AI object detection. It processes feeds from IP cameras to identify specific objects in real time using OpenCV and TensorFlow. The software operates as a server with a web interface for monitoring and configuration, allowing users to manage multiple camera streams from a single dashboard.

The application focuses on local processing to maintain privacy and reduce latency. It is typically deployed on hardware that supports AI accelerators, such as the Google Coral, to improve detection speeds and reduce CPU load. Users can define specific detection zones and masks through the web interface to ensure the system only analyzes relevant areas of the video feed, which prevents unnecessary alerts and reduces hardware strain.

Key features

• Real-time object detection using TensorFlow and OpenCV
• Integration with Home Assistant via a custom component
• Motion detection used to trigger object detection processes
• Video recording with retention policies based on detected objects
• Continuous 24/7 recording capabilities
• RTSP re-streaming to limit concurrent connections to cameras
• Low-latency live views via WebRTC and MSE
• Communication with external systems through MQTT

Frigate utilizes a multiprocessing architecture to prioritize real-time performance over the processing of every individual frame. This approach minimizes resource consumption by only analyzing video segments where motion is detected. The system includes a built-in mask and zone editor to refine detection areas and a multi-camera scrubbing tool for reviewing recorded footage. By offloading object detection to separate processes, the software maintains a higher number of frames per second during active events.

The software is designed for home automation enthusiasts and users who require a local surveillance solution with AI-driven event filtering. It integrates directly into broader infrastructure monitoring setups through its use of MQTT and specialized components for home automation platforms.

This tool provides a local infrastructure for video surveillance that emphasizes privacy and automation through local AI processing.