Owli-AI Research
Research
We investigate how visual teaching materials can become accessible and how camera images can support orientation.
Research strands
What we work on
-
01
How can circuit diagrams and charts become accessible?
We investigate representations as text, audio or tactile material. This includes the Pattern Collection for teaching materials and the funded IncluLearn.AI project.
Fig. 6 Generated UML class diagram.
Figure from Diethelm Bienhaus; Michael Kreutzer (2026).
View the figure and research paper: A Pattern Collection for Generating Accessible Teaching Materials for Blind and Visually Impaired Students in Computer Science and Electrical Engineering -
02
How can training data for sidewalk detection be generated?
The 2026 ECCV workshop contribution investigates automatic annotation of pedestrian scenes. Way-Buddy experiments with sidewalk detection and acoustic direction cues.
Fig. 2: Comparison of accepted and rejected annotations in the BGH dataset after postprocessing.
Figure from Justine Buss; Michael Kreutzer; Seyed Eghbal Ghobadi (2026).
View the figure and research paper: Open-Vocabulary Annotation for Pedestrian Scene Understanding under Domain Shift
Apps and demonstrators enable practical testing and user feedback. Their detail pages describe functions, maturity and research connections.
Research development
From bachelor's thesis to ECCV and the Innovation Prize
Justine Buss's work illustrates how a thesis can develop into a published research contribution and a visible result for Owli-AI.
-
1 / 3
Bachelor's thesis
The starting point was Justine's bachelor's thesis on assistive computer vision and pedestrian-centered scene annotation.
-
2 / 3
ECCV Workshops 2026
The results were developed into the workshop contribution “Open-Vocabulary Annotation for Pedestrian Scene Understanding under Domain Shift”.
-
3 / 3
Karl-Heinz Lust Innovation Prize
On 1 October 2026, the bachelor's thesis received the THM Karl-Heinz Lust Innovation Prize.
Projects
All projects-
IncluLearn.AI – AI-Assisted Generation of Accessible STEM Teaching Materials
IncluLearn.AI develops an AI-assisted process that semantically interprets complex visual STEM content and generates accessible text, audio, and tactile representations from it.
Funded research and transfer project
-
AHRUS Audible High Resolution Ultrasonic Sonar
AHRUS is a research prototype supporting human echolocation with audible ultrasound, described in the HCII 2018 publication.
Prototype
Publications
All publications-
A Pattern Collection for Generating Accessible Teaching Materials for Blind and Visually Impaired Students in Computer Science and Electrical Engineering
Pattern Languages of Programs, People and Practices (EuroPLoP 2025), LNCS 16494, Springer (2026) - Paper
Diethelm Bienhaus; Michael Kreutzer
-
Open-Vocabulary Annotation for Pedestrian Scene Understanding under Domain Shift
ECCV 2026 Workshops · ACVR & CDEL (2026) - Workshop contribution
Justine Buss; Michael Kreutzer; Seyed Eghbal Ghobadi
-
A Framework for Developing Modular Mobility Aids for People with Visual Impairment: An Indoor Navigation Use Case
(2023) - Paper
Florian von Zabiensky; Grigory Fridman; Oguz Ozdemir; Sebastian Reuter; Michael Kreutzer; Diethelm Bienhaus
-
An Electronic Guide Dog for the Blind based on Artificial Neural Networks
(2021) - Paper
S. Lopatin; F. v. Zabiensky; M. Kreutzer; K. Rinn; D. Bienhaus
-
Ultrasonic Waves to Support Human Echolocation
HCII (2018) - Paper
Florian von Zabiensky; Michael Kreutzer; Diethelm Bienhaus
Audio summaries
Research in audio form
These AI-generated audio summaries from NotebookLM make research projects and publications easier to access in audio form. Additional songs are available on the same page.