road brAIn

Inspiration

Ever been late for something because you were stuck in traffic? Us too, and unfortunately it's a frequent issue in highly populated areas like Atlanta. That's why we decided to create RoadBrain, to help streamline commutes for everyone.

What it does

RoadBrain uses machine learning to identify congestion on major roadways by analyzing traffic density at various locations. Our system helps civil engineers evaluate deficiencies in traffic management and rectify those issues as efficiently as possible.

How we built it

We trained the YOLO object detection model to recognize vehicles within a video and display bounding boxes around each object. We also incorporated OpenCV's KCF tracker to focus on each distinct object as it moves through each frame. The backend systems, including the algorithms that determined congested areas, were coded in Python and deployed on Google Cloud's Deep Learning Virtual Machines for optimized performance.

Challenges we ran into

Though the project presented numerous challenges, one of the greatest roadblocks was implementing an object tracker across multiple frames. Lag delays and flickering bounding boxes took considerable amounts of time to identify and solve, and we felt very gratified when the final result reached a level that we were satisfied with.

Accomplishments that we're proud of

We're most proud of how we were able to maximize the accuracy of the object detection model. By adjusting various parameters and filtering out unrelated results, we were able to pinpoint vehicles with impressive precision.

What we learned

We learned to utilize a variety of libraries and frameworks, such as numpy, and pandas, to incorporate our ML model in the most efficient way possible. We also strengthened our understanding of object detection, frame-by-frame tracking, and their countless applications in the transportation industry.

What's next for road brAIn

We hope to increase our client's productivity by bettering RoadBrain's usability, such as selecting specific roadways or pinpointing them on a map. We also envision tracking and storing data about traffic patterns, enabling urban planners to view historical trends and evaluate the performance of their transportation systems.