A control room for physical brand experiences. A queryable graph of every person, zone, and dwell in your activation. Replayable. Askable. Deliverable.
Real-time detection in your browser. See it live.
During the activation, the system captures zone entries, dwell times, and movement as structured events. Those events feed a live operator view, a 3D replay, and a client report — all from the same graph.
Real-time on-device detection in your browser today; OpenCV + YOLOv8 + ByteTrack in production. Faces never persist. Visitors are anonymous, session-scoped IDs.
Live dashboardAnonymous avatars walking a scale model of the activation. Scrub the timeline. Fly the camera. Isolate a zone. The client sees what happened, not just numbers.
Open the twinClaude or GPT-4o translates the client's question into a Cypher query against the spatial graph. Answers come back in seconds, with charts and a highlighted subgraph. No analyst required.
Ask the roomDetection emits structured events during the activation. Events accumulate in a session-scoped spatial graph. Downstream consumers — the live view, the 3D twin, and the report — read from that graph independently and in real time.
Zone entries, dwell events, and movement paths are emitted from the perception layer as structured events.
Events are written to a session graph. Every person, zone, and interaction becomes a queryable node.
The graph drives a live operator dashboard and a scrubbable 3D replay. Both surfaces read from the same store.
Ask plain-English questions. The system translates them to graph queries and returns answers with charts.
Frames are processed on-device. Raw video auto-purges in minutes; only the structured graph survives. Person IDs reset per session — we never link a visitor to themselves across days, cameras or activations.
GDPR / CCPA / UK DPA-aware by construction. Per-pixel masking available for sensitive surfaces.
A live dashboard, a 3D replay, and a client report — delivered as a measurement line item on any activation.