AI analytics
Face recognition, ANPR, PPE, safety, crowd, traffic and behaviour analytics run on your hardware — no frame leaves your network.
Face recognition, ANPR, PPE, safety, crowd, traffic and behaviour analytics run on your hardware — no frame leaves your network.
InsightFace (SCRFD + ArcFace) 512-dim embeddings. 1:1 verification and 1:N recognition across millions of identities, allow / block / watch lists, spoof heuristics and Lost & Found reunification.
YOLOv8 detectors for helmet, vest, glove, mask, safety shoes and goggles. Per-zone enforcement, industry presets and per-person compliance linked to supervisor action.
DM-Count crowd density with YOLO fallback. Density heatmaps, queue depth, dwell time, stampede early-warning, OD matrix and sector trends.
Intrusion, loitering, trip-wire, object left behind, a self-learning anomaly baseline and cross-camera re-identification with trail export.
YOLOv8n + Tesseract pipeline. India HSRP (all 39 state/UT codes + BH), EU, US and GCC plates. Hotlist alerts, e-Challan adapter, VAHAN / SARATHI / FASTag enrichment, fuzzy search and vehicle trail.
Fall detection, fight / aggression, fire & smoke, abandoned object, and YAMNet audio anomaly detection for scream, gunshot, glass break and explosion.
Wrong-way, stopped vehicle, collision, queue, congestion, pedestrian-on-road, speed estimation, lane-discipline scoring and banned-vehicle detection.
Embedded llama.cpp server with an OpenAI-compatible API, hot-swappable GGUF models and Apple Silicon Metal acceleration — generative AI that stays on-prem.
• Local CPU — ONNX Runtime, Ultralytics, InsightFace
• NVIDIA Triton for shared GPU
• DeepStream for high-density GPU
• Jetson TensorRT, Hailo-8L, Raspberry Pi TFLite
See sub-second live viewing, on-prem AI analytics and air-gap deployment in a live demo.