Cross-Camera Tracker

Cross-Camera Tracker lets operators monitor a single target individual across multiple live or recorded camera feeds simultaneously. By providing one reference photo, the system continuously matches the target's biometric identity across every connected source — even when they move between camera zones.

Use Cases

• Multi-entrance monitoring — track a VIP or flagged individual as they move through a building.
• Multi-branch coverage — watch several remote locations at once from a single control screen.
• Event security — confirm the presence of a specific person across an event venue in real time.

How It Works

1. Upload a reference photo — InsightFace ArcFace generates a 512-d embedding for the target face.
2. Connect camera sources — add multiple RTSP streams, webcam indexes, or recorded video files.
3. Start tracking — each feed is processed in parallel; whenever the target's face is detected with similarity above the configured threshold, a match event is raised.
4. View results — each camera panel highlights match frames with a confidence score and timestamp.

Configuration

Best Practices

• Use a clear, front-facing reference photo with good lighting.
• Keep the similarity threshold above 0.45 to reduce false positives across multiple feeds.
• On CPU-only systems, limit concurrent feeds to 2–3 to maintain acceptable latency.
• RTSP streams benefit from the built-in auto-reconnect — the tracker resumes automatically if a stream drops.