Wednesday, March 01, 2023

Deep Learning driven Domain Adaptation of Point Clouds

Topic and Goal of the Thesis

The training of neural networks for object detection requires a large amount of data, whose manual annotation is very time-consuming. An alternative is the use of simulation data, but these rarely manage to sufficiently represent real conditions. In this thesis, approaches based on a Generative Adversarial Network (GAN) will be investigated to improve the fidelity of automatically generated simulation data.

Working Points

• Literature review on the use of GANs for domain adaptation of point clouds
• Extension and optimization of an existing network architecture
• Training of the GAN with the help of an existing framework
• Evaluation of proposed techniques

Requirements

• Good English or German language skills
• Experience with Python
• Experience with Machine Learning is an advantage (not a must)
• Reliability, commitment and enjoyment of working independently

Note: Please attach brief resume and grade summary.