The Rainer Stiefelhagen Lab

Stiefelhagen Lab works on computer vision, multimodal interaction, and human-centered AI systems that understand people, scenes, and activities from images, video, audio, and other sensors. The group develops methods for visual recognition, activity understanding, multimodal fusion, and scene reasoning, often with deep learning and transformer-based models. A major focus is building robust datasets and benchmarks for challenging real-world settings, including chart understanding, document layout analysis, egocentric and industrial video understanding, autonomous driving, panoramic scene reasoning, sleep staging, and medical image segmentation. The lab also explores assistive technologies such as navigation support for blind and low-vision users and robotic guide-dog systems, where perception must be reliable under adverse and safety-critical conditions. Students can expect work on dataset design, multimodal learning, action segmentation, visual question answering, localization, and robustness under domain shift, label noise, and complex environments. The lab combines algorithm development with evaluation on realistic data and applications that connect AI research to healthcare, accessibility, robotics, and intelligent perception systems.