warehouse-safety

Cosmos SafetyNet: CR2-powered forklift safety reasoning with structured JSON outputs and a review dashboard.

VAST Builders Challenge: Scribner warehouse safety andon board detects near-miss hazards using Cosmos captions, YOLO11 detections, Embed1 semantic search on CoreWeave GPUs. Kubernetes deployment with live video, operator review, W&B retraining. https://luma.com/vastsf

Real-time computer vision system using YOLO11 to detect workers, track movement, and prevent forklift collisions in warehouse safety zones.

An ESP32-based indoor positioning system using trilateration for real-time collision prevention in warehouses. Includes a React web app with interactive simulation, distance visualization, and audio alarms for crash prevention demonstrations.

Warehouse safety AI that predicts forklift-pedestrian collisions before they happen. YOLOv11 → BoT-SORT → homography → custom-trained CNN (93% acc, Grad-CAM validated). Streamlit demo.

KAVACH: explainable temporal video intelligence for warehouse operations, with YOLO perception, tracking, behaviour evidence, incident replay, SQLite, and a grounded local assistant.

Ask hours of camera footage a question in plain English. You write the checks. Runs on a laptop, costs nothing.

Agentic vision AI for warehouse safety review, built on NVIDIA NIM. Nemotron VL perception + reasoning-VLM verification, with a public eval harness showing verification roughly doubles precision.