Paper accepted for publication at ECCV 2026
Our joint work with Queensland University of Technology entitled “Predictive Photometric Uncertainty in Gaussian Splatting for Novel View Synthesis” has been accepted for publication at ECCV 2026.
Osnabrück Vision AI (OSVIA) Lab
The OSVIA Lab researches methods that teach machines to see – from the basics of classical computer vision to modern deep learning approaches. Our goal is to develop algorithms that process visual data reliably, efficiently, and explainably. We work on problems ranging from the recognition and analysis of complex scenes to 3D reconstruction and robust machine learning. With our research, we bridge the gap between theory and application – for advances in science, industry, and society.
Universität Osnabrück: AG Computer Vision and Robust AI
Bergische Universität Wuppertal: AG Applied and Computational Mathematics
Recent Publications
More Publications-
Predictive Photometric Uncertainty in Gaussian Splatting for Novel View Synthesis
Galappaththige, C., J., Gottwald3, T., Stehr, P., Heinert, E., Suenderhauf1, N., Miller, D., Rottmann, M. (2026)
-
Probabilistic Label Spreading: Efficient and Consistent Estimation of Soft Labels with Epistemic Uncertainty on Graphs
Klees , J., Riedlinger, T., Stehr, P., Böddecker, B., Kondermann, D., Rottmann, M. (2026)
-
Can We Challenge Open-Vocabulary Object Detectors with Generated Content in Street Scenes?
Mütze, A., Ilyas, S., Dörpelkus, C, Rottmann, A. (2025)
Datasets
More Datasets-
Benchmark for Label Error Detection Methods in Object Detection
A Modular Framework and Benchmark for Object Detection Datasets. (2025)
-
OoDIS
Anomaly Instance Segmentation Benchmark. (2024)
-
SegmentMeIfYouCan
A Benchmark for Anomaly Segmentation. (2021)
Projects
Ongoing and recent work with our research and industry partners.
-
Project LiDiMoVer
Turning urban LiDAR data into AI-powered, searchable insights for safer and smarter mobility in Wesel.
BUW EFRE-NRW2026–2028
-
Unified Uncertainty Estimation for Fine-tuned Open-Vocabulary Models in Image Classification and Object Detection
Making modern AI systems more reliable and transparent when operating in open-world environments.
DFG UOS2026–2028
-
Project REFRAME
Investigate the possibilities and limitations of so-called foundation models in computer-aided image recognition and to develop new methods for improving these models.
BMFTR UOS2024–2027
-
Project RELiABEL
How to make the interpretation of image data for automated driving more reliable.
BMFTR UOS2024–2027
-
Just Scan It 3D
Easy creation of 3D models with photogrammetry and AI
EFRE-NRW BUW2024–2027
-
Cooperation with QUT Center for Robotics
Exploring the frontiers of foundation models in computer vision.
DAAD QUT BUW2024–2025
-
BMBF Junior Research Group UnrEAL
Making the use of artificial intelligence safer.
BMFTR BUW2022–2025 (finished)
Team
A multidisciplinary group committed to understanding and improving AI robustness in real-world scenarios.
Meet the Team
News
More News
Best Student Paper Award at VISAPP 2026
Paper
VISAPP 2026
Our doctoral student Martha Teiko T. received the Best Student Paper Award at VISAPP 2026 for the paper “FutrTrack A Camera-LiDAR Fusion Transformer for 3D Multiple Object Tracking,” co-authored by Ori Maoz.
Project LiDiMoVer
BUW
EFRE-NRW
LiDiMoVer is an innovative smart city research initiative focused on advancing urban mobility.
Paper accepted at CVPR 2026
Paper
CVPR 2026
Our paper PRIMU Uncertainty Estimation for Novel Views in Gaussian Splatting from Primitive-Based Representations of Error and Coverage” has been accepted for publication at CVPR 2026.
Inaugural lecture by Prof. Rottmann and Prof. Mossakowski
UOS
On 20.01.2026, the Institute of Computer Science celebrated "50 years of computer science at Osnabrück University". Prior to this, Prof. Rottmann and Prof. Mossakowski presented their research areas in their inaugural lectures.
Paper accepted at WACV 2026
Paper
WACV 2026
Our paper “Can We Challenge Open-Vocabulary Object Detectors with Generated Content in Street Scenes?” has been accepted to WACV 2026.