Hetereogeneous systems such as the one you could find in F1 car, generate heterogeneous dataset.
Mechanical stresses and extreme conditions during the race could corrupt these data. Futhermore some kinds of signals such as audio and video require specific data processing pipelines. Our state of art computer vision algoritmhs are able to mix this source creating predicitve models.
- On-board Camera Tagging Use Deep Learning based classifier to recognized interesting events using only data from rear/front cameras
- External Camera Tagging Use Deep Learning based classifier to recognized interesting events using only data from external cameras
- Telemetry based frame tagging Exploit telemetry data to label frames with a more accurate model.
- Real Time Frame based Ranking tagged frame to expose most informative ones using only camera data
- Predictive Frame based Ranking tagged frame to expose most informative-to-be ones using only camera data
- Real Time Telemetry based Enrich the real time ranking algorithm using features extracted from telemetry data
- Predictive anomaly detector module Enrich the prediction ranking algorithm using features extracted from telemetry data Enrich the real time ranking algorithm using features extracted from telemetry data
- 3D car models Create a 3D model of the car from camera cars to create artifical panoramic view of the car from different point of views (drones-like camera)
- Trajectories projections Project actual car trajectory over the video as an “arrow” of the direction the car is going to follow
- Real time social fresh trends Write a ranked shortlist of trendy tweets/fb posts/topics
- RT widgets Create more visual widgets exploiting new information (e.g., represent wheel speed over a 3d representation of the wheel itself)