Georg Hess, Electrical engineering

Georg Hess is an industrial PhD student at the company Zenseact and in the research group Signal processing

Discussion leader is Dr. Martin Danelljan, ETH, Zürich

Examiner is professor Tomas McKelvey and

Main supervisor is professor Lennart Svensson, Division of Signal processing and Biomedical engineering

Abstract

This work addresses challenges involved in achieving safe and efficient autonomous driving, particularly with regard to perception, which is the ability of the vehicle to sense and interpret its environment. While deep learning approaches have been successful in this area, they suffer from certain shortcomings, such as inability to estimate inherent uncertainties in perception tasks and the need for large amounts of human-annotated data. To address these issues, we propose a probabilistic model for object detection, a self-supervised pre-training scheme for Transformer-based 3D object detectors, and a model connecting automotive lidar data to natural language. These contributions aim to improve the accuracy and efficiency of autonomous driving technology.