Inspiration
Tyre degradation remains one of motorsport’s biggest unknowns. Teams spend millions on testing and still rely on guesswork during races. We wanted to create a smarter, more accessible way to understand tyre wear using AI and publicly available race footage.
As mentioned in the try it out link!
link
What it does
Project Tyre-Net uses computer vision and telemetry fusion to predict tyre degradation in real time.
It analyzes video frames lap-by-lap, compares textures, and detects subtle wear changes—turning visual data into actionable strategy insights.
How we will build it
We will train a Siamese DINOv2 transformer on tyre images synchronized with telemetry data (speed, temperature, braking).
A multimodal attention layer fuses both streams to forecast wear.
Challenges we may run into
- Synchronizing video, telemetry, and weather data with precise timestamps.
- Building a balanced dataset across compounds and lighting conditions.
- Maintaining real-time inference speed without losing model accuracy.
Accomplishments that we are aiming for
- Achieving consistent degradation predictions across unseen tracks and lighting.
- Creating one of the first open-source tyre degradation datasets for motorsport.
- Integrating AI insights into a fully functional real-time monitoring dashboard.
What we learned
- The value of data fusion — combining visual and telemetry inputs improves prediction accuracy significantly.
- Real-time deployment requires architectural optimization beyond model training.
- Small feature engineering changes can drastically affect predictive stability.
What's next for Project Tyre-Net
- Build a full working prototype for the proposed solution
- Deploy as a cloud-based analytics service for teams and broadcasters.
- Integrate sustainability metrics by quantifying tyre waste reduction per race.
Open the the try it out link for detailed presentation! link
Log in or sign up for Devpost to join the conversation.