About this Session
There are several challenges within the field of in-cabin monitoring that have to be addressed properly in order to create a well-functioning, feasible system. In this talk, the three main challenges will be laid out together with approaches to overcome them.
Different vehicle models and use cases require tailored camera configurations in terms of resolution, field of view, camera position and more. We will introduce approaches on how to determine the optimal configuration even by the time the vehicle only exists as a computer model.
Creating high-quality training data for in-cabin monitoring is a big challenge because of the uncountable possibilities of people’s actions and movements. Human centered data generation will be the core topic of this section – how to efficiently create high-quality human data.
Finally, we will dive into the specifics of algorithm development. Which algorithms and architecture are recommended to achieve high accuracy and how can you create a feasible system?
1. It is important to carefully choose the optimal camera configuration at the very beginning and tools like simulators can be used to improve the process.
2. Why data generation for in-cabin monitoring is such a challenging topic and how it can be addressed by employing different methods of real and synthetic data generation.
3. The importance of a strategic approach to algorithm development to ensure a feasible product.