-
-
Pathhole detectioin
-
Traffic Violation Detection
🚀 Inspiration Urban roads today face two major issues: poorly maintained road surfaces (like potholes) and frequent traffic rule violations that lead to accidents and congestion. Manual monitoring is slow, inconsistent, and reactive. We wanted to build a smart system that can automatically detect and report potholes and violations using real-time video feeds — making roads safer and smarter using AI.
💡 What it does Our project is an AI-based Smart Road Monitoring System that: Detects potholes using video frames and image processing. Identifies traffic violations such as helmetless riding, wrong-lane driving, and vehicle type misuse. Classifies traffic congestion in real-time (low, medium, high). Displays all this information on a user-friendly dashboard and sends alerts to authorities. In short, it replaces manual road monitoring with intelligent automation.
🛠️ How we built it We used YOLOv8 trained on custom datasets (from Roboflow) to detect potholes and violations. Built our model in Google Colab and exported the .pt file for local testing. Implemented real-time video analysis using OpenCV and Python. Developed a Flask-based backend to process camera feeds and run detection models. Designed a live dashboard UI that visualizes traffic status and violations using HTML, CSS, and JavaScript. Integrated a chatbox assistant for clients to query traffic conditions dynamically.
🧱 Challenges we ran into Dataset collection and annotation for potholes and violations was time-consuming and crucial for accuracy. Real-time video processing required optimization to reduce lag and false positives. Managing multiple class detections simultaneously in one pipeline was tricky. Integration of the YOLOv8 model with a responsive dashboard took time and testing. Local deployment using Flask + OpenCV had port and environment issues initially.
🏆 Accomplishments that we're proud of Successfully built and deployed a real-time detection system that works with local video feeds. Achieved accurate detection of both potholes and multiple traffic violations. Developed a working dashboard + assistant chatbox to interact with the system live. Trained and tested our own custom YOLOv8 model in just 36 hours.
📚 What we learned Deepened our understanding of object detection models and their deployment outside Colab. Learned how to build and train custom YOLO datasets using Roboflow. Strengthened skills in Python, Flask, and OpenCV integration. Gained experience working under hackathon pressure with real-world constraints.
🔮 What's next for AI Pothole & Traffic Violation Detector Integrate the system with city surveillance cameras and cloud infrastructure for broader deployment. Add license plate recognition (ALPR) to automatically fine violators. Enhance the dashboard with real-time maps and analytics reports. Make the system scalable for smart cities and public safety departments. Publish our dataset and model open-source for use in future civic-tech solutions.
Log in or sign up for Devpost to join the conversation.