Master's thesis: Bayesian Neural Networks for semantic bulk material inspection

Fraunhofer-Gesellschaft

ThesisGerman needed: not said

Still open on 3 Nov

You apply on the company's own website.

Key facts

Where you work: Company and home
Subject you study: Computer Science, Mathematics, Physics, Electrical Engineering and Information Technology

Analyze the state of the art: You get up to speed on current research on Bayesian Neural Networks (BNNs), uncertainty quantification and semantic segmentation.
Develop AI models: You implement and train BNN-based models for the semantic segmentation of industrial bulk materials.
Design shift detection: You develop methods to reliably detect data and concept shifts while plants are running.
Evaluate in practice: You test and validate your approaches not only in theory but directly on real sensor data from our industrial bulk material demonstrators.
Individual focus: You contribute your own ideas; we agree the exact focus of the work together based on your interests.

You are enrolled (university / university of applied sciences) and study computer science, mathematics, physics, electrical engineering and information technology or a comparable field
You have good programming skills in Python and ideally first experience with common deep learning frameworks (e.g. PyTorch).
You have great interest in topics such as computer vision, machine learning and industrial image analysis
You enjoy programming and like to learn new things
You bring scientific curiosity and look forward to technical discussions and exchange in the team.

Central location: Only a few minutes' walk from the KIT Campus Süd and very good public transport connections.
Flexible working: A mix of mobile work and work on site: by arrangement with your supervisor, but work also takes place on site at the institute.
Individual supervision: Benefit from our professional guidance and open, communicative exchange in the team.
Exciting tasks and room for ideas: A high degree of personal responsibility and the chance to actively contribute your own ideas.

Analyze the state of the art: You get up to speed on current research on Bayesian Neural Networks (BNNs), uncertainty quantification and semantic segmentation.
Develop AI models: You implement and train BNN-based models for the semantic segmentation of industrial bulk materials.
Design shift detection: You develop methods to reliably detect data and concept shifts while plants are running.
Evaluate in practice: You test and validate your approaches not only in theory but directly on real sensor data from our industrial bulk material demonstrators.
Individual focus: You contribute your own ideas; we agree the exact focus of the work together based on your interests.

You are enrolled (university / university of applied sciences) and study computer science, mathematics, physics, electrical engineering and information technology or a comparable field
You have good programming skills in Python and ideally first experience with common deep learning frameworks (e.g. PyTorch).
You have great interest in topics such as computer vision, machine learning and industrial image analysis
You enjoy programming and like to learn new things
You bring scientific curiosity and look forward to technical discussions and exchange in the team.

Central location: Only a few minutes' walk from the KIT Campus Süd and very good public transport connections.
Flexible working: A mix of mobile work and work on site: by arrangement with your supervisor, but work also takes place on site at the institute.
Individual supervision: Benefit from our professional guidance and open, communicative exchange in the team.
Exciting tasks and room for ideas: A high degree of personal responsibility and the chance to actively contribute your own ideas.