RoadGuard AI

->Inspiration What we're trying to solve

The poor physical condition of the roadways, especially when combined with an abundance of potholes, causes significant amounts of accidents, damages to vehicles, and traffic delays throughout the country of India. The current processes used by the government for identifying and reporting on the presence of such damages (due to a combination of manual inspection and citizen complaints) is inefficient and time-consuming. We felt inspired by both the problem statements related to the Smart India Hackathon and the Smart City Mission to create a solution that automates the process of identifying potholes through the use of Artificial Intelligence, allowing for faster, data-driven actions. The solution we created (RAI) RAI uses image and video data (or pictures) with the aid of computer vision software to identify potholes in the roadway. It will create a bounding box around the pothole(s), measure the size/severity of the pothole, then place a GPS location on a web-based dashboard for authorities to view, prioritize and track repair activities. Citizens will also be able to easily report road problems and have them forwarded to the authorities for action.

->The technical solution

We developed a complete web application with a back-end database containing the geographic coordinates of each detected pothole. The back-end uses the Flask and FastAPI web frameworks to process images and store results in a database using the geographic coordinates. The front-end of the application (built using React, HTML, CSS and JavaScript) allows the user to view the detected potholes on an interactive map and on an administrative dashboard.

->Obstacles we faced

•Gathering and preparing quality pothole data. •Training the AI model to recognize potholes under various lighting and roadway conditions. •Linking live detection of potholes with a web interface. •Mapping GPS coordinates accurately. •Tuning system's performance for optimal speed of pothole detection.

->Achievements we are proud of

•Accurate detection of potholes with bounding box annotations. •Severity classification system (low, medium, high). •Fully functional dashboard with real-time mapping functionality. •End-to-end pipeline flow from image upload through AI-generated results visualized.

->Experiences gained

We have gained the ability to build deep learning systems that can be integrated into full-stack web applications, utilize real-world datasets, deploy AI services, and develop systems that will support smart cities. We also gained valuable experience in teamwork, problem-solving, and task planning on a hackathon level.

->Future Plans for RoadGuard AI

•Mobile Applications to allow citizens to report potholes. •Connection with drones and CCTV Cameras. •AI to predict traffic and repairs in real time. •Large Scale Implementation for Smart Cities and Government Agencies.

Built With

• css
• github
• google-maps
• html5
• javascript
• mysql
• node.js
• opencv
• pythonwithflask
• react.js
• tensonflow
• yolov8