Maximiliane Gruber

I am working within the field of intelligent video analytics and investigate the influence of various image characteristics on machine vision systems. Such image characteristics depend on the employed image acquisition system (e.g. noise, blur, coding), and on environmental coditions (e.g. light, weather, location). The goal of my work is to reduce the “domain gap” between training and test data in these cases. In my investigations, I regard the machine vision system as black box.

Typical machine vision scenarios comprise object detection and tracking for autonomous driving, license plate recognition, and optical character recognition. Application to further domains like medical imaging are possible.

I offer theses in the field of intelligent video analytics at any time.

Master’s Theses

• “Emulation of Image Coding Artifacts Employing Adversarial Learning”
• “Emulating Complex Image Artifacts with Unpaired Image-to-Image Translation Methods”
• “Simulation and Learning of Depth of Focus Blur for Domain Adaptation”

Bachelor’s Theses

• “Document Image Quality Assessment for Text Recognition Systems”
• “Emulation of Image Artifacts by Learning in the Frequency Domain”
• “Setup of a demonstrator for depth estimation employing a stereo camera and a LIDAR”

Research Internships and Project Theses

• “Image quality analysis from decoded video frames”
• “Modeling of environmental influences in Blender using Python”
• “Assessment of the Influence of Video Coding on Text Recognition”
• “Investigation of the Domain Gap Induced by Different Data Sets in Machine Vision for Autonomous Driving”
• “Evaluation of Bi-Directional Object Tracking”
• “Benchmarking the Influence of Video Distortions on Object Tracking”