Loading…
Loading…
Catch the events nobody thought to set up. Visylix learns each camera's normal and alerts the moment something is out of the ordinary, with no rules, zones or labels to write.
Visylix Anomaly Detection learns each camera's normal: the usual movement, objects and activity for that view, at each time of day. After a short learning period it watches continuously and alerts on anything that does not fit, from an out-of-place object to unusual activity in a yard that is normally still. It keeps adapting as the scene changes, never stops watching while it learns, and gets sharper each time an operator marks a false alarm. It complements the rule-based analytics rather than replacing them: intrusion and line crossing catch what you planned for, and anomaly detection catches what you did not. Every anomaly can open an incident, notify a supervisor or trigger a talkdown warning through a Sudarshan rule.
What it catches, how it cuts false alarms, and what happens next.
Builds its own picture of normal for every camera, unique to that scene.
Nothing to draw or configure, so it protects scenes that are hard to describe with fixed rules.
Flags out-of-place objects and unusual activity that no rule was set up to look for.
Learns what is normal at different times of day, so nightfall does not trigger alarms.
Keeps adapting as the scene changes and never stops detecting while it learns.
When an operator marks an alert as a false alarm, that camera learns from it.
Choose how unusual activity must be before it alerts, balancing early warning against noise.
Open an incident, notify a supervisor or play a talkdown warning through Sudarshan rules.
Built for the problems security, safety and operations teams face every day.
Watch fences, yards and remote locations where the unexpected is hard to define with fixed rules.
Surface any out-of-pattern activity in offices, warehouses and shops after closing.
Add a self-learning layer to substations, water works and data centres beyond fixed rules.
Keep watch where no operator is looking at the feed, and alert someone the moment something changes.
Visylix watches, decides and acts, all on your own hardware.
Runs on the cameras you choose, with zones per camera and per model. Each camera learns its own normal, and marking a false alarm tightens thresholds automatically, never below your limit.
Sudarshan rules combine this analytic with others using AND/OR logic, time sequences and access-control events. Test on real history first, with version history, snooze, hourly caps and maintenance windows.
Notifications, alarm incidents in the operator console, signed webhooks, MQTT, SIEM and PSIM, talkdown through the camera speaker, siren and door-strike relays, and escalation when a critical alert is not acknowledged.
Search every detection in plain English, Hindi or Hinglish. Radha, the on-premise AI copilot with 111 tools, asks before any high-impact action and logs every call.
Combine analytics in a single Sudarshan rule for layered coverage.
AI analytics are a separate subscription, billed per camera per month on the cameras you choose, with any of the 22. Talk to us for AI pricing and a demo on your own footage.
Visylix learns each camera's normal: the usual movement, objects and activity for that view, at each time of day. Once it has learned, it alerts on anything out of the ordinary. There are no rules, zones or labels to set up.
No. Each camera learns what is normal at different times of day, so nightfall and a quiet weekend are not treated as unusual. When an operator marks an alert as a false alarm, that camera learns from it.
It works alongside them. Intrusion, line crossing and the other analytics catch the events you planned for; anomaly detection catches the ones nobody thought to set up. Both can drive the same Sudarshan rules.