VIGIL — AI Pedestrian Safety System

Overview

VIGIL (Vehicle-Installed Guard for Injury Limitation) is a production-grade AI safety system that mounts on warehouse forklifts to detect pedestrians in real time. It ships with 7 swappable AI models — from lightweight YOLO11 Nano to high-accuracy RT-DETR transformers.

The Challenge

Warehouse forklifts operate in dynamic environments with pedestrians, requiring real-time detection with zero tolerance for missed detections while maintaining low latency for immediate alerts.

The Solution

Built a full-stack system with a FastAPI backend running a multi-model detection pipeline (YOLOv8n through RT-DETR), custom safety zone engine, gRPC streaming for real-time alerts, and a React command-center dashboard with live KPI monitoring.

Key Results

• 7 swappable AI models with zero-downtime runtime switching
• Real-time MJPEG streaming with HUD overlays and detection counts
• Custom safety zone engine with violation tracking and alerting
• Full analytics suite with hourly charts, camera distribution, and PDF report generation
• gRPC + Kafka integration for enterprise-scale event streaming