Stefan Gumhold
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Stefan Gumhold has a total of 102 co-authors. Here are the top co-authors by number of shared publications:
- Raimund Dachselt
Technische Universität Dresden
5 shared publications - Nishant Kumar
Technische Universität Dresden
4 shared publications - Benjamin Russig
Computer Graphics and Visualization, TUD Dresden University of Technology, Dresden, Germany
3 shared publications - Joachim Staib
TU Dresden, Nöthnitzer Straße 46, Institute of Software and Multimedia Technology, Dresden, Germany
3 shared publications - Marc Satkowski
Fraunhofer Institute for Process Engineering and Packaging IVV, Dresden, Germany
3 shared publications - Rufat Rzayev
Interactive Media Lab Dresden, TUD Dresden University of Technology, Dresden, Germany
3 shared publications
Latest Publications
- Open Access
- 25 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Comparing Ground-Anchored Augmented Reality Visualizations of Waste Heat Redistribution
Urban waste heat from IT infrastructure is increasingly relevant for district heating, but remains largely invisible to non-experts. Situated augmented reality (AR) can make redistribution more legible by placing conceptual routes between source and receiver buildings. We compare three outdoor AR representations: a cutaway pipe (Pipe), floating ribbon (Ribbon), and ground-conforming trace (Trace). In a within-subject study (N = 24), all three enabled accurate receiver-building identification, indicating basic spatial legibility. Subjective and comparative judgments differed more strongly: Pipe was clearer than Trace and ranked highest for perceived support of understanding; Trace was least intrusive but weaker for communicative suitability, especially understanding; Ribbon was intermediate mainly for understanding. These findings highlight a trade-off between explanatory explicitness and visual intrusion in public-facing AR visualizations of urban waste heat redistribution.
- 25 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Open AccessComparing Ground-Anchored Augmented Reality Visualizations of Waste Heat Redistribution
Urban waste heat from IT infrastructure is increasingly relevant for district heating, but remains largely invisible to non-experts. Situated augmented reality (AR) can make redistribution more legible by placing conceptual routes between source and receiver buildings. We compare three outdoor AR representations: a cutaway pipe (Pipe), floating ribbon (Ribbon), and ground-conforming trace (Trace). In a within-subject study (N = 24), all three enabled accurate receiver-building identification, indicating basic spatial legibility. Subjective and comparative judgments differed more strongly: Pipe was clearer than Trace and ranked highest for perceived support of understanding; Trace was least intrusive but weaker for communicative suitability, especially understanding; Ribbon was intermediate mainly for understanding. These findings highlight a trade-off between explanatory explicitness and visual intrusion in public-facing AR visualizations of urban waste heat redistribution.
- 2 August 2026
- Published by Springer Nature
- p. 156-169
Designing Digital Humans: A Systematic Review of User Experience in Avatar-Mediated Interaction
- 2 August 2026
- Published by Springer Nature
- p. 156-169
Designing Digital Humans: A Systematic Review of User Experience in Avatar-Mediated Interaction
- Open Access
- 17 June 2026
- Published by Wiley in Computer Graphics Forum
Tubes or Ribbons? Comparing Texture‐space Visualization for Multivariate Line Data
Multivariate line data is critical for analyzing flow fields, agent systems, and dynamic trajectories. Embedding secondary variables along spatial paths using surface‐based primitives such as ribbons and circular tubes introduces challenges related to perspective, scale, and distortion. Perceptual trade‐offs due to these challenges remain unclear. We address this gap through a controlled user study with 10 experts in computational fluid dynamics and visualization, performing four identical analysis tasks involving both spatial (requiring location‐based relationships) and non‐spatial (requiring attribute value comparisons) aspects. The tasks were performed using a prototype with interactivity limited to controlling the camera. While quantitative measures showed no significant performance differences between embedded visualizations on ribbons and tubes, we found a clear subjective preference for tubes among the study participants. Furthermore, their feedback indicates surface‐based embeddings are generally helpful and should be utilized more often.
- 17 June 2026
- Published by Wiley in Computer Graphics Forum
Open AccessTubes or Ribbons? Comparing Texture‐space Visualization for Multivariate Line Data
Multivariate line data is critical for analyzing flow fields, agent systems, and dynamic trajectories. Embedding secondary variables along spatial paths using surface‐based primitives such as ribbons and circular tubes introduces challenges related to perspective, scale, and distortion. Perceptual trade‐offs due to these challenges remain unclear. We address this gap through a controlled user study with 10 experts in computational fluid dynamics and visualization, performing four identical analysis tasks involving both spatial (requiring location‐based relationships) and non‐spatial (requiring attribute value comparisons) aspects. The tasks were performed using a prototype with interactivity limited to controlling the camera. While quantitative measures showed no significant performance differences between embedded visualizations on ribbons and tubes, we found a clear subjective preference for tubes among the study participants. Furthermore, their feedback indicates surface‐based embeddings are generally helpful and should be utilized more often.
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-22
Mixed Presence in Mixed Reality: Charting the Challenges and Opportunities
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-22
Open AccessMixed Presence in Mixed Reality: Charting the Challenges and Opportunities
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-19
Beyond Links: Exploring Visual Representations of Multi-View Relations in Mixed Reality
This paper investigates associations, explicit representations of relations between multiple views in Mixed Reality (MR). While research on 2D desktop environments offers extensive recommendations for communicating relations between multiple views, MR environments lack such systematic guidance, necessitating adapted solutions that consider their spatial affordances. To address this gap, we systematically explored association techniques in existing research. Building on established 2D multi-view literature and refining insights from prior design principles, we developed a codebook to describe view relations and their representations. Applying it to a corpus of 44 immersive multi-view approaches, we identified recurring design strategies and synthesized them into a design space of visual association techniques adapted for immersive contexts. Based on a lightweight prototyping framework, we validate the utility of the design space through three envisioning scenarios, demonstrating how associations can support exploration, coordination, and sensemaking in MR applications. Our results inform the design of MR multi-view environments.
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-19
Open AccessBeyond Links: Exploring Visual Representations of Multi-View Relations in Mixed Reality
This paper investigates associations, explicit representations of relations between multiple views in Mixed Reality (MR). While research on 2D desktop environments offers extensive recommendations for communicating relations between multiple views, MR environments lack such systematic guidance, necessitating adapted solutions that consider their spatial affordances. To address this gap, we systematically explored association techniques in existing research. Building on established 2D multi-view literature and refining insights from prior design principles, we developed a codebook to describe view relations and their representations. Applying it to a corpus of 44 immersive multi-view approaches, we identified recurring design strategies and synthesized them into a design space of visual association techniques adapted for immersive contexts. Based on a lightweight prototyping framework, we validate the utility of the design space through three envisioning scenarios, demonstrating how associations can support exploration, coordination, and sensemaking in MR applications. Our results inform the design of MR multi-view environments.
- Open Access
- 12 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
IGUANA: Immersive Guidance, Navigation, and Control for Consumer UAV
- 12 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessIGUANA: Immersive Guidance, Navigation, and Control for Consumer UAV
- Open Access
- 8 August 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Computer Graphics and Interactive Techniques
- Vol. 8 (4) , 1-21
Fast and Accurate Parameter Conversion for Parametric Human Body Models 54
Parametric human body models, especially from the family of SMPL, are widely used in fields such as motion capture, 3D human pose and shape estimation, as well as for animation or synthesis of human motion. A problem that arises when different body model types are to be used in the same workflow is that model parameters are not interchangeable. Existing methods to convert parameters are based on optimization. While these methods achieve low conversion errors, they exhibit long runtimes, making them suboptimal for the conversion of real-time data or large datasets. In this paper, we improve on these aspects by presenting a deep learning-based conversion approach that is both accurate and fast, enabling the interplay of different body model types in real-time scenarios. To this end, we use a fully-connected neural network to jointly convert translation, shape, and pose parameters for different SMPL-based body models and show that we can accurately convert sequences from different datasets. Our method is able to provide feedback on its conversion quality, which can be used to enhance the trustworthiness of a superordinate processing pipeline. We also implement a demo application for the real-time streaming and visualization of digital human twins. We use our conversion method to transform parameters of different body model types to a common representation before streaming and show that we can do so in real-time even for sequences that are running at 120 Hz.
- 8 August 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Computer Graphics and Interactive Techniques
- Vol. 8 (4) , 1-21
Open AccessFast and Accurate Parameter Conversion for Parametric Human Body Models 54
Parametric human body models, especially from the family of SMPL, are widely used in fields such as motion capture, 3D human pose and shape estimation, as well as for animation or synthesis of human motion. A problem that arises when different body model types are to be used in the same workflow is that model parameters are not interchangeable. Existing methods to convert parameters are based on optimization. While these methods achieve low conversion errors, they exhibit long runtimes, making them suboptimal for the conversion of real-time data or large datasets. In this paper, we improve on these aspects by presenting a deep learning-based conversion approach that is both accurate and fast, enabling the interplay of different body model types in real-time scenarios. To this end, we use a fully-connected neural network to jointly convert translation, shape, and pose parameters for different SMPL-based body models and show that we can accurately convert sequences from different datasets. Our method is able to provide feedback on its conversion quality, which can be used to enhance the trustworthiness of a superordinate processing pipeline. We also implement a demo application for the real-time streaming and visualization of digital human twins. We use our conversion method to transform parameters of different body model types to a common representation before streaming and show that we can do so in real-time even for sequences that are running at 120 Hz.
- 27 February 2025
- Published by Springer Nature
- p. 358-370
Enhanced Plant Phenotyping Through Spatio-Temporal Point Cloud Registration
- 27 February 2025
- Published by Springer Nature
- p. 358-370
Enhanced Plant Phenotyping Through Spatio-Temporal Point Cloud Registration
- Open Access
- 24 March 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 38 (19) , 21610-21618
Quantile-Based Maximum Likelihood Training for Outlier Detection
Discriminative learning effectively predicts true object class for image classification. However, it often results in false positives for outliers, posing critical concerns in applications like autonomous driving and video surveillance systems. Previous attempts to address this challenge involved training image classifiers through contrastive learning using actual outlier data or synthesizing outliers for self-supervised learning. Furthermore, unsupervised generative modeling of inliers in pixel space has shown limited success for outlier detection. In this work, we introduce a quantile-based maximum likelihood objective for learning the inlier distribution to improve the outlier separation during inference. Our approach fits a normalizing flow to pre-trained discriminative features and detects the outliers according to the evaluated log-likelihood. The experimental evaluation demonstrates the effectiveness of our method as it surpasses the performance of the state-of-the-art unsupervised methods for outlier detection. The results are also competitive compared with a recent self-supervised approach for outlier detection. Our work allows to reduce dependency on well-sampled negative training data, which is especially important for domains like medical diagnostics or remote sensing.
- 24 March 2024
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 38 (19) , 21610-21618
Open AccessQuantile-Based Maximum Likelihood Training for Outlier Detection
Discriminative learning effectively predicts true object class for image classification. However, it often results in false positives for outliers, posing critical concerns in applications like autonomous driving and video surveillance systems. Previous attempts to address this challenge involved training image classifiers through contrastive learning using actual outlier data or synthesizing outliers for self-supervised learning. Furthermore, unsupervised generative modeling of inliers in pixel space has shown limited success for outlier detection. In this work, we introduce a quantile-based maximum likelihood objective for learning the inlier distribution to improve the outlier separation during inference. Our approach fits a normalizing flow to pre-trained discriminative features and detects the outliers according to the evaluated log-likelihood. The experimental evaluation demonstrates the effectiveness of our method as it surpasses the performance of the state-of-the-art unsupervised methods for outlier detection. The results are also competitive compared with a recent self-supervised approach for outlier detection. Our work allows to reduce dependency on well-sampled negative training data, which is especially important for domains like medical diagnostics or remote sensing.