Technische Universität Braunschweig – Institute for Communications Technology (IfN) (Subcontractor)
The Institute for Communications Technology (IfN) with its Chair "Signal Processing and Machine Learning" and the affiliated Deep Learning Lab (TUBS.dll) headed by Prof. Dr.-Ing. Tim Fingscheidt has been researching with a focus on perception in the field of machine learning on image and audio/speech data for decades. In the area of speech and audio processing in automotive applications, the Chair is nationally leading, has won numerous prizes (two ITG prizes, Technology Transfer Prize, national and international Student Best Paper Awards) and is among international leaders.
In machine learning for perception processes, the Chair is currently focussing on the following fields of research: image compression, semantic segmentation, corner case detection, robustness via teacher-student approaches, and unsupervised approaches for domain adaptation. In the latter two fields, a CVPR Workshop Best Paper Award was awarded in 2019 and 2020, respectively. IfN was already involved in the project KI Plattform, contributes to KI Validation as a subcontractor and in KI Data Tooling as a full university member.
Within the project KI Delta Learning, TU Braunschweig is investigating, how domain adaptation can be achieved through unsupervised learning. In the area of continuous learning, the additional challenge is to investigate to what extent unsupervised online domain adaptation can be executed in the vehicle without the neural network "forgetting".