Institution profile

Technical University of Munich

Academic institutioneurope · de
Official website
Research library3,045linked papers
Opportunities0open roles
Selected work

Representative Papers

TEyeD: Over 20 Million Real-World Eye Images with Pupil, Eyelid, and Iris 2D and 3D Segmentations, 2D and 3D Landmarks, 3D Eyeball, Gaze Vector, and Eye Movement Types

Feb 03, 2021International Symposium on Mixed and Augmented Reality

Existing eye movement and gaze estimation research is hindered by the scarcity of large-scale, multi-scenario, high-precision publicly available datasets—especially in real-world VR/AR environments. To address this, we introduce the largest head-mounted device-collected eye image dataset to date (>20 million images), spanning diverse daily activities and VR/AR scenarios. It features the first multi-device synchronized acquisition and unified annotation of comprehensive eye-related attributes: 2D/3D eye landmarks, pupil/iris/eyelid segmentation masks, parametric 3D eyeball models, gaze vectors, and fine-grained eye movement types. We propose a geometrically constrained eyeball fitting and gaze estimation method, integrated with a semi-automatic labeling pipeline validated by domain experts. This dataset establishes the first real-world benchmark for eye movement analysis, yielding consistent improvements of 12–28% in eye movement estimation and gaze prediction accuracy across multiple state-of-the-art models.

71 citations5 influentialRead paper

Benchmarking the CoW with the TopCoW Challenge: Topology-Aware Anatomical Segmentation of the Circle of Willis for CTA and MRA

Dec 29, 2023arXiv.org

The Circle of Willis (CoW) suffers from a scarcity of high-quality voxel-level annotations in CTA/MRA imaging, reliance on labor-intensive expert manual segmentation, and poor guarantee of topological consistency. Method: We introduce the first publicly available voxel-level multi-class CoW dataset—comprising 13 vascular structures with paired MRA/CTA volumes—and propose a topology-aware segmentation framework: (i) a novel VR-assisted annotation paradigm ensuring anatomical plausibility; (ii) a multimodal registration and topology-constrained segmentation network; and (iii) topology-sensitive metrics including branch F1 and topo-Dice. Contribution/Results: This benchmark has attracted >140 teams across four continents. State-of-the-art models achieve ≈90% Dice on most arterial branches, while exposing persistent topological matching bottlenecks—particularly for communicating arteries and anatomical variants.

24 citations2 influentialRead paper

Eye-tracked Virtual Reality: A Comprehensive Survey on Methods and Privacy Challenges

May 23, 2023arXiv.org

This paper addresses privacy leakage risks arising from the correlation between eye-tracking data and visual stimuli in VR environments. We systematically survey full-stack VR eye-tracking technologies—from pupil detection and gaze estimation to cognitive modeling—published between 2012 and 2022, alongside their associated privacy threats. First, we establish the first cross-disciplinary survey framework bridging VR eye-tracking and privacy protection, identifying three privacy-centric research directions. Second, we propose a novel co-design paradigm integrating eye movement authentication with data anonymization, synergizing computer vision, human-computer interaction modeling, differential privacy, adversarial generation, and biometric encryption. Third, we clarify the technological evolution trajectory and privacy threat landscape, and introduce quantifiable evaluation metrics and an implementable defense roadmap. Our work provides both theoretical foundations and practical guidelines for developing secure and trustworthy VR systems. (149 words)

21 citationsRead paper

Bridging Language and Action: A Survey of Language-Conditioned Robot Manipulation

Dec 17, 2023

This work addresses the semantic gap between natural language instructions and robotic physical actions to enhance the naturalness and reliability of human-robot collaboration. We propose the first four-dimensional taxonomy for language-conditioned robotic manipulation—comprising reward shaping, policy learning, neurosymbolic AI, and foundation model–driven approaches—and systematically analyze their fundamental limitations in generalization and safety. Integrating large language models (LLMs), vision-language models (VLMs), neurosymbolic reasoning, and multimodal semantic parsing, we develop a unified analytical framework spanning semantic extraction, environmental assessment, and auxiliary task design. Our analysis rigorously characterizes the performance boundaries of each paradigm for the first time, establishing theoretical foundations and concrete technical pathways toward safe, generalizable, and interpretable language-driven robotic systems.

10 citationsRead paper

Analyzing the Impact of Simulation Fidelity on the Evaluation of Autonomous Driving Motion Control

Jun 02, 20242024 IEEE Intelligent Vehicles Symposium (IV)

This study addresses the challenge of inconsistent evaluation of autonomous driving control algorithms due to varying fidelity levels in vehicle dynamics models. The authors develop a high-fidelity vehicle model compatible with Autoware and systematically derive a hierarchy of reduced-fidelity models through controlled simplifications. Leveraging over 550 simulations and real-world track data—encompassing speeds up to 267 kph and lateral accelerations of 15 m/s²—they quantitatively assess, for the first time, how simulation fidelity impacts trajectory tracking performance. Furthermore, they propose an application-oriented model simplification criterion based on acceleration margin, which explicitly defines acceptable levels of model reduction under different acceleration limits. This framework provides a principled basis for selecting appropriate model fidelity when evaluating control algorithms in simulation.

9 citationsRead paper
Recent publications

Latest Papers

Hardware-Free Robotics Laboratories in Mixed Reality

Sep 16, 2026

"This study addresses the limitations of traditional robotics education, which heavily relies on screen-based simulations and lacks practical experience in real-world environments. To tackle this issue, the project introduces MR-Robotics LAB, a mixed reality platform that facilitates the generation of trajectories via MATLAB and their realistic-scale reproduction on the Meta Quest 3, complete with collision detection and grasping functionalities. The platform leverages MATLAB, JSON, and Unity for data exchange and visualization, offering a seamless transition from simulation to mixed reality. Preliminary evaluations indicate that students find the setup easy to use and helpful for understanding workspace concepts, with an average rating of 4.56 out of 5, and 83% of them expressing willingness to utilize the platform in introductory courses."

0 citationsRead paper