Complex video analysis,
simplified by AI.
Transform standard perspective CCTV and traffic cameras into time-synchronized 2D top-down plan trajectories. Follow full entrance-to-exit desire lines, detect loitering, and converse with your scene using grounded natural language AI.
Why stacked video + 2D plan solves surveillance
Video alone hides perspective and direction. 2D maps alone lose visual context. Zenvisia synchronizes both with a grounded AI intelligence layer.
Planar Homography Projection
Maps perspective pixel coordinates to calibrated ground-level metric coordinates. North is truly North, and speed is computed in meters-per-second, not camera pixels.
OpenStreetMap CalibratedFull Entrance-to-Exit Tracks
No short fading tails. The trajectory line begins at the exact entrance corridor used (North, South, East, West) and tracks all the way to where the subject currently stands or leaves.
End-to-End Desire LinesGrounded AI Query Engine
Ask questions instead of scrubbing hours of footage. Answers cite real counts, timestamps, peak time intervals, and IDs that instantly highlight on the map.
Zero HallucinationsPrivacy-First Architecture
No facial recognition, license plate extraction, or biometric identity profiling. Every target receives an anonymous sequential integer (`#1`, `#2`, `#12`) compliant with strict GDPR standards.
100% AnonymizedWhat you can ask Zenvisia
Click on any sample question to see how Zenvisia answers with grounded trajectory numbers:
47 people entered from the north between 19:00 and 20:00 yesterday (Mon 31 Aug). Peak was 19:22–19:28 (12 entries), consistent with a film letting out at the multiplex.
Ready to analyze your camera footage?
Experience the full interactive demo with real Aupark Plaza video and 2D trajectories, or upload your own footage for GPU processing.