RoadSense AI is an end-to-end Computer Vision and telemetry application designed to detect, log, and map road damage in real-time. Powered by a custom-trained YOLO architecture, the system accurately identifies road hazards such as potholes and surface cracks, dynamically evaluates an overall Road Health Score, logs the localized GPS coordinates, and plots incident markers instantly onto an interactive map.
- Dual Execution Modes: *
📸 Static Image Upload:Drop road snapshots to audit individual sections.🎥 Live Camera Scan:Real-time, continuous video analysis using integrated device cameras or dashcams via WebRTC streaming.
- Real-Time Object Detection: Fine-tuned computer vision layer tracking road structural failure anomalies (Potholes, Longitudinal, Transverse, and Alligator Cracks).
- Automated Telemetry Matrix: Logs timestamps, failure types, hardware confidence scores, and coordinate matrices synchronously.
- Interactive Spatio-Temporal Map: Live-rendered geographical layout utilizing Folium engine to drop visual incident hazard tags.
- Automated Quality Scorecard: Real-time road health metric calculations dropping down from a baseline score of 100 based on hazard frequency and severity.
- Frontend Dashboard Interface: Streamlit Engine
- Core Computer Vision Pipeline: Ultralytics YOLOv12
- Geospatial Mapping Utilities: Folium & Streamlit-Folium Wrappers
- Video Analytics Pipeline: Streamlit-Webrtc Framework & PyAV Media Library
- Structured Data Telemetry Engine: Pandas Framework
RoadSense-AI/
├── backend/
│ ├── main.py
│ ├── webcam.py
│ └── yolo12s_RDD2022_best.pt # Fine-tuned YOLO weights
├── frontend/
│ └── app.py # Multi-mode Streamlit Dashboard app
├── requirements.txt # Deployment environment dependencies
└── README.md # Project documentation