Speaker Details

Speaker Company

Tommy Johansson

With 17 years of experience, Johansson has navigated the realm of advanced driver assistance systems (ADAS) across OEMs, T1 suppliers and presently at Kognic, an automotive industry ML tool supplier. His expertise spans various disciplines throughout the AD/ADAS development chain. Johansson is particularly passionate about tracking, delving into intricate data fusion algorithms and ensuring functional safety within these systems. Throughout his career, he has contributed significantly to advancing the efficacy and reliability of AD/ADAS technologies. His journey has encompassed hands-on involvement across the development spectrum, allowing him to gain a holistic understanding of the intricacies involved in creating robust and innovative AD/ADAS solutions.

Presentation

The significance of iterative understanding of ML data sets

Machine learning (ML) models are only as good as the data sets they are trained on. The quality of the data set plays a pivotal role in determining the performance and reliability of the resulting models. However, achieving high-quality data sets is not a one-time task; it requires an iterative process of understanding, assessing and refining data to enhance model performance continually. This presentation delves into the critical role of iteratively assessing data set quality concerning model performance in the realm of machine learning, focusing on ADAS/AD use cases.

The audience will learn:

  • Iterative data set quality: the necessity of ongoing assessment for improving model performance in machine learning
  • The data set-model relationship: how data set attributes directly impact model outputs, revealing biases and flaws
  • Continuous improvement: feedback loops must be established between data set assessment and model performance for ongoing enhancements
  • Real-life applications: practical examples showcasing iterative data set analysis' impact on actionable decisions
  • Informed decision making: using data set insights for adjusting model architecture and data collection strategies