BIT Technology Solutions GmbH is located in the heart of Munich. Our goal is to make autonomous driving safe and reliable with synthetic data for training and validation of artificial intelligence and machine learning.
Collaboration is at the heart of our success. Our international team works together to realize an extraordinary vision that results in innovative software and scalable solutions for car manufacturers like BMW and their suppliers.
We also have a strong focus on research projects.
To grow our development team we look forward for motivated
for a master’s thesis in a rapidly growing and promising field of autonomous driving. You should have a background and practical experience in computer vision and deep learning, besides programming skills in Python and ideally C.
Deep neural networks have shown to be a great success in computer vision algorithms for driver assistance systems. However, the safety of such systems is a challenging factor, as they have shown to have various weaknesses. Adversarial examples are a phenomenon of neural networks which are considered as their Achilles’ heel. A perturbed image (adversarial image), that might look identical to the unperturbed image for the human observer, can lead to completely different predictions of a neural network. The topic of this Master‘s Thesis requires to train a semantic segmentation network on street scene images. The master student should perform an extensive literature research in order to summarize recent methods for generating adversarial examples. Afterwards, some of those methods should be implemented in order to trick the trained network on these manipulated images. In the end methods to efficiently defend against these so called adversarial attacks should be utilized.
Start date: as soon as possible
Sounds interesting? Please send your application.