To be filled: 01.12.2018
Audi has long been a driving force in the area of highly-automated driving, and has repeatedly documented its progress in this technology. With the Audi AI traffic jam pilot, they have presented the world's first system with SAE level 3 conditional automation. Already existing driver assistance functions are able to support the driver in different driving tasks, increasing comfort and safety on public roads.
Those systems require an accurate representation of the car's surroundings. Information about the environment is extracted by various sensors (e.g. radar, lidar, camera). Reliable camera systems with accurate and efficient image processing algorithms are indispensable for the detection of lane markings, traffic signs and other road users.
Domain adaptation with Generative Adversarial Networks (GANs) allows to transform the domain of images, e.g. illumination and weather conditions, while preserving authentic appearance. The models trained may be used for data augmentation increasing robustness of machine learning techniques or for domain alignment of images captured at different environmental conditions (e.g. day and night) by connected cars.
The objective of this thesis is to develop artificial intelligence methods for training and evaluation of domain adaptation GANs including the following tasks:
Diese Stelle ist bei der AUDI AG in Ingolstadt zu besetzen
Reference code: I-A-55473
Questions answered by Herr Michael Hofweber
by calling +49(0)841-89-38804