SignPatrol

Kartenansicht
Listenansicht
Meldeformular
Mobile Ansicht
Hardwareprototyp

Inspiration

Our inspiration for SignPatrol stemmed from the realization that dirty or unreadable street signs not only make navigation difficult but also pose significant safety risks for drivers, pedestrians, and cyclists alike. In bustling cities and growing suburbs, keeping up with maintenance is crucial yet challenging. We wanted to utilize cutting-edge technology to address this often-overlooked issue.

What it does

SignPatrol is a system that leverages artificial intelligence to detect dirty, vandalized, or obstructed street signs. It features a mobile application designed to allow citizens to report signs as well as a robust system equipped with camera mounted on a vehicle. As the vehicle moves, the camera captures images of street signs, which are then analyzed by our AI model to determine their condition. Each sign is cataloged with a description and rated on a scale from 0 to 100, indicating its cleanliness and visibility.

How we built it

We built the project using state-of-the-art artificial intelligence coupled with commodity hardware, enabling prototyping at a pace that would otherwise not have been possible. Leveraging existing frameworks allowed us to deliver an accurate classification within the short timeframe of a hackathon. An intuitive dashboard allows stakeholders to review uploaded data and view the associated cleanliness rating and the content description.

Challenges we ran into

Gathering a comprehensive dataset posed a significant hurdle. Creating high-quality images of street signs in varying conditions was surprisingly tough, as they are mostly located in remote locations. We needed pictures of wear, vandalism, and obstruction. Additionally, integrating external libraries and tools proved resource-intensive.

Accomplishments that we're proud of

The way SignPatrol turned out to be a functioning prototype in a remarkably short timeframe. The performance of this prototype exceeded our expectations, showcasing a high level of accuracy in sign recognition and analysis. This success lays a strong foundation for the future of SignPatrol.

What we learned

Throughout this project, we've gained invaluable insights into the importance of training set creation, data engineering and knowledge of local geography. We've also learned about the effectiveness of leveraging existing ecosystems to create minimal viable products in a short timeframe.

What's next for SignPatrol

We are planning to partner with local governments to bring our detection system to streets everywere. Additionally, our system will soon spot graffiti and other damages, improving the aesthetics of cities. An additional benefit of our system will be the contribution of more accurate street sign data to OpenStreetMap, bringing the benefit of high-quality data to everyone.

In collaboration with local governments, SignPatrol will deploy our advanced system to streamline street sign maintenance. The scheduling interface will feature automated cost and workforce calculations for repairs, efficient routing for maintenance tasks, and streamlined permission management for maintenance-related activities, ensuring a smooth and compliant operational flow.