Pothole Detection with AI
Turning Bumps into Bytes.
The problem Pothole Detection with AI solves
Potholes and damaged roads are a major cause of accidents, vehicle damage, and traffic congestion. Currently, road maintenance depends on manual inspection or citizen complaints, which is slow, inefficient, and lacks transparency. Proposed Solution RoadSafeAI is an AI-powered pothole detection and reporting system that uses computer vision + GPS to automatically identify road damages and notify municipal authorities in real-time. How It Works Detection – AI model scans images/video from smartphone cameras, dashcams, or CCTV to detect potholes/cracks. Geo-Tagging – Each pothole is marked with GPS coordinates and severity level. Reporting – Data is sent to a central dashboard for authorities with photos + location. Monitoring & Feedback – Dashboard shows a heatmap of road damage; citizens get notified when repairs are done. Impact Faster Road Repair → Reduces accidents and traffic issues. Smart City Integration → Builds a transparent road maintenance system. Low Cost → Uses existing smartphones & cameras (no special equipment needed). Crowdsourced AI → Citizens, cabs, and buses contribute data passively. Tech Stack Computer Vision: TensorFlow/Keras, YOLOv8 (for pothole detection). Mobile App: React Native (camera + GPS integration). Backend: Flask/Node.js + Firebase/MongoDB. Dashboard: Web app with Google Maps API / Leaflet.js for heatmaps.
Challenges we ran into
While building the RoadSafeAI system, we faced a problem when trying to upload images along with location data from the frontend (React app) to the backend server (FastAPI). The upload would fail with an error like "Failed to fetch," meaning the frontend could not communicate with the backend successfully.
This was frustrating because we could not send pothole images and their GPS locations for detection, which is the core functionality of the project.
How We Solved It:
We first made sure the backend server was running properly without any errors.
We added special settings (called CORS middleware) in the backend to let the frontend connect to it without being blocked by security restrictions.
We installed all necessary Python packages, especially one called python-multipart, required for handling file uploads properly.
We double-checked that the frontend was sending data to the exact correct web address of the backend.
After these fixes, uploads worked perfectly, and the system could process images and locations as expected.
