Team Idea
SentinelEdge is a local-first, low-latency Physical AI inspection system combining unsupervised visual defect detection with closed-loop telemetry actions.
The pipeline ingests edge video feeds via OpenCV, running a PatchCore model trained via Intel Anomalib and converted to Intel OpenVINO Intermediate Representation (IR, FP16/INT8) for high-throughput edge CPU inference. An integrated lightweight agent supervisor parses anomaly heatmaps, assesses structural severity thresholds, and dispatches deterministic hardware override commands with sub-50ms inference latency.
Stack: Python, Intel OpenVINO, Anomalib (PyTorch), OpenCV, Streamlit, FastAPI.
Team Members
Shambhu Shekhar Sinha
shambhu_shekhar_sinha
