Thesis: Bridging the Sim2Real gap for sensor data in an industrial context via generative models* in Waldkirch bei SICK AG

SICK AG

SICK is one of the world’s leading solutions providers for sensor-based applications in the industrial sector. Founded in 1946 by Dr.-Ing. e. h. Erwin Sick, the company with headquarters in Waldkirch, Breisgau near Freiburg ranks among the technological market leaders. With more than 50 subsidiaries and equity investments as well as numerous agencies, SICK maintains a presence around the globe. SICK has more than 12,000 employees worldwide and generated a group revenue of around EUR 2.3 billion in the 2023 fiscal year.

Thesis: Bridging the Sim2Real gap for sensor data in an industrial context via generative models*

SICK AG

Waldkirch (bei Freiburg), DE, 79183

Full-time

Winter semester 2024/25 - limited to 5-6 months

Machine learning (ML) models are increasingly hungry for data. In industrial contexts, high-quality labeled data is a scarce and costly resource. Synthetic data enables ML model training, algorithm evaluation, or sensor design without the burden of exhaustive data collection. However, generating realistic data is difficult, because not all attributes of sensors or the environment are captured (i.e., simulation-reality-gap). Recently, generative models have emerged as a promising avenue, potentially helping in generating synthetic data that closely mimics real-world scenarios.
The goal in this thesis will be to evaluate the potential of generative models to bridge the gap between realistic and synthetic sensor data in an industrial context.

YOUR TASKS:

You perform a literature review on generating realistic sensor data via generative models
You familiarize yourself with sensor simulation in Nvidia Omniverse
You explore the applicability of 1-2 generative approaches on a specific industrial use case
You evaluate the performance of considered approaches on a downstream task

YOUR PROFILE:

You are familiar with generative models such as diffusion models or GANs
You may bring experience in machine learning domains such as computer vision, audio processing, or NLP, demonstrating expertise in system training, evaluation, and understanding of these systems
You might have designed your own games or tinkered with the industrial Metaverse
You are passionate about exploring the applicability of Generative AI for a specific industrial use case

YOUR APPLICATION:

We are looking forward to your online application
Sarah Disch
Job-ID 35903
All applications will be treated confidentially

At SICK, we see people, not gender.

We put great emphasis on diversity, reject discrimination and do not think in categories such as gender, ethnicity, religion, disability, age or sexual identity.