Your cameras, your server, no cloud subscription

What it does Viseron turns IP cameras into a monitored security system. It records video, detects motion, recognizes faces and license plates, and identifies objects — all processed locally. The project wraps together an NVR backend with a web interface for configuration and viewing.

The interesting bit The “components” architecture lets you mix and match capabilities: YOLO for object detection, Google Coral or CUDA for hardware acceleration, specific camera protocols. It is essentially a curated integration layer for open-source computer-vision tools, packaged so you do not have to wire them together yourself.

Key highlights

• Runs fully offline; no cloud dependency or subscription
• Hardware acceleration via CUDA, Google Coral Edge TPU, or OpenCL
• Supports RTSP streams, motion detection, face recognition, and license-plate recognition
• Web-based configuration UI (no manual YAML editing required)
• Component system allows swapping detectors, cameras, and notifiers

Caveats

• README is thin on technical specifics; hardware requirements, performance benchmarks, and supported model versions are not stated
• “Local only” is the pitch, but the degree of phone-home from bundled dependencies is unclear from the sources

Verdict Worth evaluating if you want a Frigate alternative with a web UI and broader component swap-ability. Skip if you need detailed documentation upfront or a project with transparent performance numbers.