SecureVu delivers real-time AI object detection, smart recording, and full control over your surveillance system — entirely on your own hardware. Private by design.
SecureVu transforms any IP camera into an AI-powered sentry. Scroll to see the system in action.
Pull the SecureVu Docker image. Up and running on any Linux server in under two minutes.
Add RTSP stream URLs in the YAML config. Any ONVIF or RTSP-compatible camera works instantly.
Realtime object detection at sub-second latency. Bounding boxes, confidence scores, and event clips.
Live feeds, timeline playback, and smart notifications — all accessible from the web UI on any device.
# config.yml cameras: front_door: ffmpeg: inputs: - path: rtsp://192.168.1.10/stream roles: [detect, record] detect: objects: [person, car] parking: ffmpeg: inputs: - path: rtsp://192.168.1.11/stream detect: objects: [car, truck] lpr: true perimeter: ffmpeg: inputs: - path: rtsp://192.168.1.12/stream motion: threshold: 25
Built on proven open-source technology, SecureVu combines AI object detection with a professional VMS in a single Docker container.
Detect people, vehicles, animals, and more in real-time using AI models running entirely on your hardware. No cloud processing, no latency.
Record only when objects are detected to save storage. Continuous recording with object-based clipping gives you full coverage without the waste.
Native support for NVIDIA TensorRT, Intel OpenVINO, Google Coral, AMD ROCm, Hailo, Rockchip NPU, and more. Use your existing hardware.
WebRTC, MSE, and JSMPEG live streams. RTSP restreaming for third-party integrations. Low-latency viewing on any device.
All processing happens locally. No footage, metadata, or events are ever sent to external servers. Your data stays yours.
Browse recordings on an interactive timeline. Search events by object type, camera, or time range with a clean, responsive web UI.
Native Home Assistant integration via MQTT. Get camera entities, motion sensors, and object detection events in your automation flows.
Automatically read and log license plates from cameras pointed at driveways, parking lots, or entrances.
Full Pan-Tilt-Zoom control with autotracking. SecureVu can automatically follow detected objects across the frame.
Connects to vision-language models to automatically analyze, summarize, and describe security events with real context.
Find past events using natural language ("person with a yellow vest") or image references. Jina CLIP v2 support.
Specialized models adaptable to your use case: wildlife species, binary states (door open/closed), or person attributes.
Granular control with custom polygons, motion masks, and PTZ auto-tracking to filter false positives and track objects.
From entry-level edge devices to high-end GPUs — SecureVu supports a wide range of AI accelerators out of the box.
Maximum GPU performance with TRT-optimized model inference
SupportedOptimized inference on Intel CPUs, GPUs, and VPUs
SupportedEdge TPU for ultra-low power object detection
SupportedHigh-performance inference on AMD Radeon GPU hardware
SupportedDedicated AI processor for embedded deployments
SupportedRK3588 & RK3566 NPU for ARM-based SBCs
SupportedNano, Orin, AGX — full Jetson platform support
SupportedAstra platform for smart camera deployments
SupportedNext-gen in-memory computing AI accelerator
SupportedCPU inference on any Linux server, no GPU required
SupportedA single Docker Compose file is all you need. SecureVu handles the rest — detection zones, recording, streaming, and alerts.
docker compose up commandSecureVu integrates with the tools and platforms you already use for home automation, alerting, and monitoring.
We run a live instance with demo cameras so you can explore the full platform — no install required.
Log in with the credentials below to explore live camera feeds, AI detections, the event timeline, and all settings.
Full documentation, configuration reference, and deployment guides at docs.secure.vu