Automated Road Surface Damage Detection System
An accuracy-focused system for businesses and municipalities that uses AI to detect road cracks and potholes from images and video, visualizing them on a map with GPS coordinates.
Duration
Contract development
Team
1-2 people
01 // OVERVIEW
Project Overview
A business-oriented web system that, when images or video are uploaded, uses AI to detect road damage and visualize it on a map with GPS coordinates. It classifies longitudinal, transverse, and alligator cracking as well as potholes, supporting the digitization of inspection work.
02 // KEY FEATURES
Key Features
Classification of Damage Types
Detects and classifies multiple damage types, including longitudinal, transverse, and alligator cracking, as well as potholes.
Map Visualization With GPS Coordinates
Extracts GPS data from image EXIF metadata and visualizes detection results on a Leaflet and OpenStreetMap map. Also includes a statistics dashboard.
Resident API With On-Demand Inference
A resident API managed by PM2 handles intake and data management, while inference is started only when called, conserving resources.
03 // TECHNICAL DEEP-DIVE
Technical Highlights
Solution
Ensuring Detection Accuracy That Holds Up on Real Footage
To avoid the trap of models that perform well on cropped images but break down on real footage, we thoroughly reviewed labels visually before training, prioritizing accuracy on real-world operational data.
04 // IMPACT & RESULTS
Results
Multi-Class
Damage types classified
GPS-Linked
Automatically mapped