Tim Gerrits

Research Projects

Uncertain Mode Surfaces in 3D Symmetric Second-Order Tensor Field Ensembles

2025 IEEE VIS

To be published

• Download Preprint

NHR Young Researcher Exchange Program

Funded Project Collaboration between NHR and JHPCN

3-Month Research Stay at the University of Tokyo, Japan

DeepSwitch - A Web-based Tool for the Introduction to Visual Analysis of Spatiotemporal Processes in Oceanographic Data

Visualization in Environmental Sciences 2025 (EnvirVis 2025), EuroVis 2025

• Show Bibtex
• Download Paper

Interactive Streaming of 3D Scenes to Mobile Devices using Dual-Layer Image Warping and Loop-based Depth Reconstruction

Journal of WSCG, 2025

• Show Bibtex
• Download Paper

Enabling Modular In-situ Workflows through CatalystMaestro and CatalystComposer

E.C. (eds) High Performance Computing. ISC High Performance 2025. Lecture Notes in Computer Science

• Show Bibtex
• Download Paper

Exploring Uncertainty Visualization for Degenerate Tensors in 3D Symmetric Second-Order Tensor Field Ensembles

2024 IEEE Workshop on Uncertainty Visualization: Applications, Techniques, Software, and Decision Frameworks

Best Paper Award

• Show Bibtex
• Download Paper
• Download Preprint

Region-based Visualization in Hierarchically Clustered Ensemble Volumes

The Symposium on Vision, Modeling, and Visualization (VMV), 2024

• Show Bibtex
• Download Paper

DaVE - A Curated Database of Visualization Examples

IEEE VIS 2024

• Show Bibtex
• Download Paper
• Download Preprint
• Visit Dave Website

InsitUE - Enabling Hybrid In-situ Visualizations through Unreal Engine and Catalyst

Springer Lecture Notes (7th International Workshop on In Situ Visualization, ISC 2024)

• Show Bibtex
• Download Paper

Game Engines for Immersive Visualization: Using Unreal Engine Beyond Entertainment

PRESENCE: Virtual and Augmented Reality

• Show Bibtex
• Download Paper
• Download Preprint

On the Computation of User Placements for Virtual Formation Adjustments during Group Navigation

IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), 2024

• Show Bibtex
• Download Preprint
• Download Paper

IntenSelect+: Enhancing Score-Based Selection in Virtual Reality

IEEE Transactions on Visualization and Computer Graphics, 2024

• Show Bibtex
• Download Preprint
• Download Paper

IEEE SciVis Contest 2023 - Dataset of Neuronal Network Simulations of the Human Brain

IEEE SciVis Contest 2023, 2023

Data Publication

• Show Bibtex
• Zenodo Access

DaVE - A curated Database of Visualization Examples

NHR Conference, 2023

Project Poster

• Show Bibtex
• Zenodo Access
• Download Poster

Leveraging BC6H Texture Compression and Filtering for Efficient Vector Field Visualization

The Symposium on Vision, Modeling, and Visualization (VMV), 2023

• Show Bibtex
• Download Paper

A Case Study on Providing Accessbility-Focused In-Transit Architectures for Neural Network Simulation and Analysis

Springer Lecture Notes (6th International Workshop on In Situ Visualization, ISC 2023)

• Show Bibtex
• Journal

Insite Pipeline - A Pipeline Enabling In-Transit Processing for Arbor, NEST and TVB

Human Brain Project Summit, 2023

Research Poster

• Show Bibtex
• Zenodo Access
• Download Poster

A Case Study on Providing Immersive Visualization for Neuronal Network Data Using COTS Soft- and Hardware

IEEE VR Workshop on Immersive Visualization Laboratories - Past, Present and Future, 2023

• Show Bibtex
• Journal

DaVE – A curated Database of Visualization Examples

NHR Future Project

Accepted project proposal for one year.

• NHR Website

MODE: A modern ordinary differential equation solver for C++ and CUDA

AIP Conference Proceedings

Accepted at the International Conference Of Numerical Analysis And Applied Mathematics (ICNAAM22), published in the 2023 proceedings.

• Show Bibtex
• Access Paper
• Access Github Code

A C++20 Interface for MPI 4.0

The International Conference for High Performance Computing, Networking, Storage, and Analysis (SC22), 2022

Research Poster

• Access Github Code
• Download Abstract
• Download Poster

Astray: A Performance-Portable Geodesic Ray Tracer

The Symposium on Vision, Modeling, and Visualization (VMV), 2022

• Show Bibtex
• Journal

Studying the Effect of Tissue Properties on Radiofrequency Ablation by Visual Simulation Ensemble Analysis

12th Eurographics Workshop on Visual Computing for Biology and Medicine (EG VCBM), 2022

Honorable Mention Full Paper Award

• Show Bibtex
• Journal

Performance Assessment of Diffusive Load Balancing for Distributed Particle Advection

Journal of WSCG, 2022

• Show Bibtex
• Download Paper

Multi-faceted Visual Analysis of Oceanographic Simulation Ensemble Data

IEEE Computer Graphics and Applications, 2022

Publication based on the winning entry of the SciVis Contest at the IEEE VIS 2020 conference.

• Show Bibtex
• Journal

Insite: A Pipeline Enabling In-Transit Visualization and Analysis for Neuronal Network Simulations

Springer Lecture Notes (6th International Workshop on In Situ Visualization, ISC 2022)

• Show Bibtex
• Journal

Bringing Visualization to National HPC Infrastructure:
A Prologue

6th International Workshop on In Situ Visualization, ISC 2022

Keynote Presentation

• Show Bibtex
• Zenodo Access
• Download Slides

PhD. Publications

Visualization of Second-Order Tensor Data and Vector Field Ensembles

Dissertation

• Show Bibtex
• Library Entry

Towards Glyphs for Uncertain Symmetric Second-Order Tensors

Computer Graphics Forum (Proc. EuroVis), 2019

Assuming Gaussian Uncertainty, This Work Proposes an Extension to Known Glyph Constructions. The Uncertainty is Represented as a Scalar Field That Lives On the Glyph Surface Which Allows for a Complete and Unique Encoding.

• Show Bibtex
• Download Paper
• Watch Video

An Approximate Parallel Vectors Operator for Multiple Vector Fields

Computer Graphics Forum (Proc. EuroVis), 2018

The Approximate Parallel Vectors (APV) Operator is a Generic Feature Extraction Method for Vector Field Ensembles. It Extracts Features at Locations, Where All Input Vectors are Approximately Parallel by Generating Two New Vector Fields From All Input Fields and Applying a Parallel Vectors (PV) Operator.

• Show Bibtex
• Download Paper

On-The-Fly Tracking of Flame Surfaces for the Visual Analysis of Combustion Processes

Computer Graphics Forum, 2018

• Show Bibtex
• Download Paper

Glyphs for Space-Time Jacobians of Time-Dependent Vector Fields

Journal of WSCG, 2017

As Jacobians are a Special Case of General Second-Order 2D and 3D Tensors, an Extension to the 2D and 3D Tensor Glyph Design is Introduced. These Glyphs Are Capable to Additionally Encode the Temporal Information Given by a Time-Dependent Jacobian by Chosing An Appropriate Mapping From an Higher Dimension Into the Visualization Subspace.

• Show Bibtex
• Download Paper
• Watch Video

Glyphs for General Second-Order 2D and 3D Tensors

IEEE Transactions on Visualization and Computer Graphics
(Proc. IEEE Scientific Visualization 2016)

A New Construction For Tensor Glyphs That are Able to Represent General Second-Order 2D and 3D Tensors. This Includes Non-Symmetric Tensors Where Eigenvectors are Non-Orthogonal or Eigenvalues are Complex Valued. These Glyphs are Constructed Using Piecewise Rational Bezier Curves and Surfaces Following A Strict Set of Properties.

• Show Bibtex
• Download Paper
• Watch Video

Selected Student Projects

Master Thesis

Eulerian Flow Field Reconstruction From Lagrangian Trajectories

A flow field can be described in two distinct ways: The Eulerian and the Lagrangian de- scription. Particle Tracking Velocimetry (PTV) for example is a measurement technique that results in trajectory data in Lagrangian description. Several numerical methods however de- mand the field to be described in Eulereian terms, such as velocity vectors on structured grids. In this thesis, a technique to reconstruct an Eulerian flow field by means of a combi- nation of local bilinear interpolation and regularized least squares, that incorporates model knowledge and spatio-temporal coherence is developed. Further, a framework is introduced to test the approach on different versions of flow fields.

GIST Feature Extraction From
Shallow Convolutional Neural Networks

Despite its complexity, Convolutional Neural Networks (CNNs) have grown to become a leading technology in object recognition for research and industry purposes. This is achieved by using huge datasets like Imagenet and long training periods of network models with several hidden layers to generate high level features for classification. However, these so called deep networks were found to be inefficient for scene classification so other databases like the Places database were introduced to train networks for this purpose. The high complexity of common CNN topologies as well as the huge size of training sets for generating features renders CNNs to be impractical. In this paper, we present our work on generating two-class scene classification features with fast and simple CNNs. To do so, we implemented different network models with varying but reduced complexity, trained on the same dataset and compared the results in terms of accuracy and time used while increasing the complexity gradually. We were able to show, that a shallow network can compete with state-of-the-art deep networks or hand-generated features. This was done in the context of a scientific team project supervised by Prof. Tönnies at the Otto-von-Guericke University of Magdeburg.

• Download Paper

Semi-automatic Catheter Reconstruction

High-dose-rate (HDR) brachytherapy is a type of internal radiation therapy that delivers high doses of radiation from implants placed close to, or inside, the tumor(s) in the body. To generate a computer-assisted radiation plan, it is necessary to segment the catheters placed in the body. In this project, we implemented and evaluated a semi-automatic algorithm to segment a catheter using an oriented LoG-Filter. The basic approach was to determine the main direction of the catheter and use this information to align the LoG-Filter. Additional filtering steps improved the results. With this simple method, we obtained results of 0.9 overlap per catheter.

The code was was written in MATLAB and applied to Computed Tomography scans.

• Download Paper (German)

Face Biometrics

The project aimed at producing an application that would allow for face detection as well as recoginition. Several different approaches have been implemented to compare results and performance. This included Eigenfaces, Fisherfaces and Local Binary Patterns Histogram. Further, a dataset for different face recognition tasks has been introduced as well as a realtime face, gender and liveliness detection.

The code was written in C++. The GUI was realized with the QT framework and the computer vision algorithms were provided by OpenCV.

Bachelor Thesis

Untersuchung von Methoden zum automatischen Einbetten fotorealistischer Objekte in Szenen ohne zusätzlicher Szeneninformationen

This thesis analyzes the realistic integration of objects into scenes. Commonly used methods, as well as their conditions and restrictions are explained, followed by the question, whether an automatic embedding process can be realized, even though the scene information is restricted to one low dynamic range image only. Therefore, different methods in the field of Style Transfers are discussed and implemented, as they try to transfer image characterstics from one image to another. An additional, own approache is then introduced, which uses an automatic white balancing algorithm to transfer illumination information from the scene to and embedded object. The results are rated in a user study about the visual quality of the computed images. The results show, that none of the used methods produces very realistic integrations for all scenes tested right away, but offer potential for further investigation.

Marker Based Augmented Reality

The project aimed at producing an application that would allow realtime rendering of 3D objects on 2D markers in a live video stream. Different types and numbers of markers as well as uploading .obj files of choice were supported.

The code was written in C++. The GUI was realized with the QT framework, marker detection and recognition was implemented using OpenCV and rendering was done by the use of OpenGL.

Movement Detection

The project aimed at producing an application that would detect moving elements in image series. Several datasets were given, including image sets with static as well as noisy backgrounds.

The code was written in C++. The GUI was realized with the QT framework and image processing supported by OpenCV.

• Download Sourcecode
• Download Paper (German)

Realtime Object Detection

The project aimed at producing an application that would allow for the realtime detection of static objects in camera streams. Several different approaches of object detection have been implemented to compare results and performance. This included color detection, feature detection as well as template matching.

The code was written in C++. The GUI was realized with the QT framework and the computer vision algorithms were provided by OpenCV.

• Download Sourcecode
• Download Paper (German)
• Download Presentation (German)

Short Movie Production - "Traumjob"

Final project for the subject "Einführung in die Allgemeine Pädagogik". The movie was produced over several weeks and included writing, filming, cutting and marketing. It was awarded the Golden Kangaroo and the Audience Award at the Magdeburg University Film Festival Videoexpo as well as first price at StudiMovie Magdeburg.

The full movie can be found here:
AG2 Traumjob