ClearPath AI

ClearPath AI can make any car a smart car by adding top-notch safety features with a click of a button. Our app is powered by AI to alert you of cars or pedestrians in your field of view, saving lives

Comment

Inspiration

Have you ever been seconds away from a crash? Your heart rate races and your life flashes before your eyes. These days car accidents are all too common and you oftentimes end up paying for mistakes that aren't even your fault. But now we have a way for everyone to get the most upgraded safety features on their existing car with just the click of a button (and no additional cost), saving lives in those crucial moments before a crash.

What it does

Our app is powered by the latest AI and computer vision to detect pedestrians and cars and output auditory, visual, and haptic warnings about obstacles in your path. It can add the latest safety features to any car, regardless of age, make, or model.

How we built it

We used YOLO with a Roboflow dataset to train an object detection and segmentation model. This model was then integrated into our Swift application to provide real-time predictions on objects in your path and immediately alert users of potential hazards. The UI was designed in FIMA and deployed with SwiftUI

Challenges we ran into

The biggest challenge we faced was integrating the AI model into the Swift application. CoreML is no longer available on Windows, so we faced some difficulty with converting the YOLO model to an Apple computer vision model. Additionally, we had some difficulty finding a good dataset to train our model on. We were eventually able to find a good pre-annotated dataset on Roboflow.

Accomplishments that we're proud of

We have a working prototype that works in real-time. We have a large potential for scalability.

What we learned

Ideate quickly, and leverage resources that you're familiar with and are already available to you

What's next for ClearPath AI

Implement crash detection based on accelerometer data from the phone, implement GPS tracking to adjust danger thresholds according to speed information, transfer the compute onto the cloud to reduce hardware requirements, extrapolate the model to identify more hazards and in different weather conditions, improve the app to provide additional feedback modes

Built With

Share this project:

Updates