CollisionGuard AI
Real-time collision detection using YOLOv8, MiDaS depth estimation, and 0–100 risk scoring.
Tech Stack
YOLOv8MiDaSPyTorchOpenCVFastAPIReactTypeScriptSupabase
Project Overview
Engineered a real-time collision detection system using YOLOv8 for object detection and MiDaS for depth estimation. Built an IoU-based motion tracking engine with a multi-factor 0–100 risk scoring system. Deployed via FastAPI backend with WebSocket streaming and a React/TypeScript frontend backed by Supabase.
Key Features
Real-time object detection using YOLOv8
MiDaS monocular depth estimation
IoU-based multi-object motion tracking
Multi-factor 0–100 collision risk scoring engine
FastAPI backend with WebSocket live streaming
React + TypeScript frontend with Supabase integration