Inspiration
Poor road conditions and undetected potholes are a major cause of accidents, vehicle damage, and traffic congestion. Despite their impact on public safety and economic efficiency, road inspection and maintenance are still largely reactive and manual in many regions. We were inspired to build a system that could automatically monitor road health in real time, reduce human effort, and provide actionable data to authorities for faster maintenance decisions. Our goal was to move beyond simple detection and build a smart infrastructure solution that integrates artificial intelligence, sensors, and cloud technologies.
What it does
RoadGuard AI is an intelligent road monitoring and maintenance system. It:Uses cameras mounted on government police or highway patrol vehicles to monitor roads while performing routine duties. Detects potholes and road cracks in real-time using a YOLO-based computer vision model. Geotags each detected defect with GPS coordinates for precise location tracking. Classifies nearby traffic into heavy, medium, and light vehicles to predict roads at higher risk of future damage. Integrates weather forecast data (e.g., rainfall, hail, storms) to prioritize maintenance of vulnerable road sections. Generates a maintenance priority score, enabling authorities to act on the most critical road issues first.
How we built it
Hardware: Dashcams with GPS feed.
AI Detection: YOLO model trained to detect potholes and cracks. Vehicle detection model to classify traffic type and volume.
Data Fusion: Combines AI detection outputs, vehicle traffic data, and weather forecast data. Calculates a priority index for maintenance based on severity, traffic load, and environmental risk.
Dashboard: Provides an interactive dashboard for authorities to view hotspots and track maintenance status.
Workflow: Vehicle captures video and GPS location. Edge AI detects potholes and cracks. Data sent to the cloud for analysis and priority scoring. Dashboard displays actionable insights for maintenance teams.
Challenges we ran into
Integrating multiple data sources (AI detection, traffic type, and weather) into a single priority metric. Real-time processing on vehicle-mounted edge devices required model optimization.
Accomplishments that we're proud of
Built a scalable system using existing government vehicles, avoiding the need for dedicated inspection fleets. Achieved accurate pothole and crack detection using YOLO models. Developed predictive maintenance logic based on vehicle load and weather conditions.
What we learned
Gained hands-on experience in edge AI deployment, cloud integration, and full-stack dashboard development.
What's next for RoadGuard AI
Try to make something which will be able to maintain the roads and not just monitor it. RoadGuard AI can be enhanced with MEMS sensors, temperature sensors, and strain gauges embedded in pavements and bridges, allowing continuous monitoring and predictive analysis of road conditions to prioritize maintenance before visible damage occurs.
Built With
- amazon-web-services
- mongodb
- npm
- python
- redis
- tailwind
- uvicorn
- vite
- yolo