Tata Elxsi has built an industrial monitoring system that turns existing camera feeds into near-real-time safety alerts. IRIS is intended for factories, warehouses and chemical facilities where human operators cannot continuously watch hundreds of screens.
Edge servers connected to the camera network sample two to five frames per second, filter for motion and run an initial lightweight model. Only safety-relevant frames and structured metadata move to AWS, rather than every stream being uploaded continuously. Cloud computer-vision models then detect hazards, while a correlation layer combines observations into higher-confidence events before notifying staff.
The design addresses privacy, bandwidth and data-residency concerns. The described deployment uses AWS’s Mumbai region for Indian facilities, with GPU-equipped edge hardware such as Nvidia Jetson AGX Orin. It can monitor conditions such as access to restricted areas or missing protective equipment, depending on the trained detectors.
Tata Elxsi says conventional review can take 15 to 45 minutes and that IRIS detects unsafe conditions within seconds. Those figures are presented by the vendor, not from an independent safety study. Automated alerts supplement workplace programs; missed detections, false alarms and site-specific camera conditions still require human testing and oversight.