Data Fusion for a new era of Telemetry
artea.com developed a ML project for a major supplier of high-tech products and systems in the automotive industry, particularly for racing cars.
The different devices and systems on a Formula 1 car generate heterogeneous datasets. Mechanical stress and extreme conditions of use during races can deteriorate data. Additionally, certain types of signals (such as audio and video) require specific processing.
Computational visualization algorithms can process heterogeneous data sources and create predictive models. The design of our platform supports race officials with an all-in-one application that allows them to monitor a plethora of variables and events.
- On-board Camera Tagging Use Deep Learning-based Classifier to recognize relevant events based on data from rear/front cameras only.
- External Camera Tagging Use Deep Learning-based Classifier to recognize relevant events using data from external cameras only.
- Telemetry-based frame tagging Leverage telemetry data to label frames in a more accurate model.
- Real-Time Frame-based Rank tagged frames to expose the most relevant based on camera data only.
- Predictive Frame based Rank tagged frames to expose useful information based on predictive models.
- Real-Time Telemetry-based Enrich real-time classification algorithms using capabilities from telemetry data.
- Predictive anomaly detector module Enrich predictive classification algorithms using capabilities from telemetry data.
- 3D car models Create a 3D model based on camera views from different standpoints (drone-like camera).
- Trajectories projections Display car trajectory on video as an arrow pointing to the current direction.
- Real-time social fresh trends Write a ranked shortlist of trendy tweets, Facebook Posts and topics.
- RT widgets Create multiple visual widgets that leverage new information (e.g., display the speed of the wheel in a 3D model).