Joint PhD Degree Programme NTUTU Darmstadt




This PhD programme is a joint collaboration between Nanyang Technological University (NTU, Singapore) and Technische Universität Darmstadt (TU Darmstadt, Germany). The programme offers opportunities to promising scientists and scholars to foster interactions between the three world class Universities in the area of interactive digital media. The principal areas of focus are Real-time Rendering, Virtual & Augmented Reality, Visual Analytics, Medical Computing,  Haptics, and Human-Computer Interaction.  The 4-years scholarships provided jointly by NTU and Fraunhofer Singapore assume that the applicants become full-time PhD students of NTU and TU Darmstadt while having two supervisors - one from NTU and a co-supervisor from TU Darmstadt. The NTU students will spent one year of their project in TU Darmstadt in the co-supervisor's lab. TU Darmstadt students will be supported by TU Darmstadt PhD scholarships and also will need to spend one year at NTU in the co-supervisor's lab. The degree of Ph.D. with mention of this collaboration between NTU and TU Darmstadt will be delivered by the academic authorities of the institution where the student has successfully defended his/her oral defense. The Doctoral degree transcript will mention the name of the student, the thesis or research title, the doctoral degree specialization or discipline, the fact that there was an international joint supervision, the date of the oral defence.




NTU-based students:


Student: Mr. Zhengkun Yi , started in January 2013, 1 year attachment at TUD: March 2015 - April 2016. (Scholarship from Fraunhofer Singapore). Degree defended 17.11.2017.

NTU supervisor: Yilei ZHANG

TU Darmstadt co-supervisor: Jan PETERS

Topic: Biomimetic Tactile Sensing and Statistical Interpretation.

Abstract: The objective of the joint degrees Ph.D. project is to develop capabilities in biomimetic tactile signal generation and interpretation for IDM, including design and develop biomimetic tactile sensor as well as tactile signal modelling and interpretation with statistical machine learning. Just like sight, hearing, taste and smell, touch is widely utilized to interact with the surrounding environments. The sense of touch contributes substantially to the recognition of the shape and size of objects, and enables us to discriminate between surface textures, recognize and manipulate objects, etc., all of which have a variety of applications in robotics, minimal invasive surgery and manufacturing industries.


Student: Ms. Zhang Xingzi,started in August 2013, 1 year attachment at TUD: September 2015 - August 2016. (NTU Scholarship). Oral presentation 5.06.2018.

NTU supervisor: Alexei SOURIN

TU Darmstadt co-supervisor: Michael GOESELE

Topic: Image-driven Haptic Interaction

Abstract: We will replace visual rendering of 3D scenes with merely displaying their images while simulating haptic interaction with the scenes displayed in images as if they were 3D models provided for such interaction. These can be also images of real scenes (photographs or video) which photorealistic 3D modelling can be a challenge.  The haptic effects can be both derived from the image as well as obtained from the optional haptic models augmenting the image. These additional invisible models will be topologically collocated with the objects displayed in the image, as well as they will be able to provide haptic effects for the simulated phenomena which do not have any visual shape.


Student: Mr. Cui Jian, started in August 2013, 1 year attachment at TUD: September 2015 - August 2016. (NTU Scholarship).

NTU supervisor: Alexei SOURIN

TU Darmstadt co-supervisor: Dieter FELLNER

Topic: Hand-controlled Shape Modeling

Abstract: Using different hand tracking devices, to design and implement a set of robust and efficient interactive hand gestures suitable for various virtual engineering designs (virtual prototyping) and crafts (freeform shape modelling). To achieve multimodal hand tracking and interaction so that it will become possible to start modelling with one type of device and then seamlessly continue it using other type of hand tracking device. In contrast to commonly used polygon and voxel based modelling as well as various surface splines, to define models of the shapes to be created and manipulated with bare hands using mathematical functions (procedures) to achieve any desired level of detail and to be able to exchange models of the shapes across the internet since mathematical functions stored by their coefficients and parameters occupy very little space.


Student: Mr. Cheng Wentao, started in August 2013,1 year attachment at TUD: March 2016 - April 2017. (Scholarship from Fraunhofer Singapore)

NTU supervisor: LIN Weisi

TU Darmstadt co-supervisor: Michael GOESELE

Topic: Human Perception Models for Real-Time Graphic Rendering

Abstract: How to speed up the rendering process for real-time computer graphics and animation is an important research problem, especially with the emerging scenario of mobile computing, where computation power and battery are scarce resources. To this aim, adaptive sampling can be performed, by varying the number of samples in the spatial-temporal light field according to certain criteria for visual contents (such as contrast of graphics in the existing algorithms) so that fewer samples are needed without quality loss of rendered quality image


Student: Mr. Huynh Ngoc Anh, started in January 2014, 1 year attachment at TUD: March 2016 - April 2017. (NTU Scholarship) Thesis pending examination.

NTU supervisor: NG Wee Keong

TU Darmstadt co-supervisor: Dieter FELLNER, Joern KOHLHAMMER

Topic: Visual Analytics for Parameter Space Exploration in Cyber-Security

Abstract: This project is on Visual Analytics exploring the combination of visualization and automated methods for an integrated analysis of massive amounts of data. One major challenge in these areas is the consideration of interactivity on the visualization side while combining and exploring automated computations.  A promising approach in visual analytics is the computation of the best possible result within a given timeframe and to gradually improve the result from this stage, if the user is interested in a particular area. While there are some first approaches, much work still has to be done for data sets that feature non-normally distributed data, skewed data distribution or other specific characteristic to ensure good results even on a first sampling. Especially for high-dimensional data sets, the parameter space exploration of automated algorithms and the analysis of the corresponding results of the applied algorithms are of high interest in many application areas, like medicine, finance, or cyber-security.  The combination of complexity, size, and velocity (the rate of new data added to the analysis) on the data side, and the need for expert support in cyber-security makes this area predestined for a visual analytics approach.


Student: Ms. Ma Jingting, started in August 2014, current 1 year attachment at TUD: September 2016 - August 2017. (Scholarship from Fraunhofer Singapore)

NTU supervisor: LIN Feng, Marius ERDT (Fraunhofer IDM@NTU)

TU Darmstadt co-supervisor: Dieter FELLNER, Stefan WESARG (Fraunhofer IGD)

Topic: Self-learning shape recognition in medical images

Abstract: In modern clinical routine, massive amounts of medical image data from 3D acquisition techniques like Computed Tomography or Magnetic Resonance Imaging are created every day. Analysing, interpreting, and processing this data automatically is useful to support the physician in diagnosis or operation planning. Model based approaches are frequently used to process medical images, e.g. for segmentation or registration tasks. The built models are usually based on manually labelled training images. In order to address the significant anatomical shape variability of certain anatomical structures, often many training data sets are needed. Though the image data is available in the clinics, it can be infeasible in practice to manually label all anatomical structures in a sufficiently large subset of these images for training. However, a lot of prior knowledge is inherent in medical images that can be learned in an unsupervised way since often no major scale, translation, rotation, and appearance differences between images of the same anatomical structure are present. The goal of this project is to exploit this prior knowledge in order to build shape models automatically from unlabeled training images. Therefore, new methods for unsupervised shape recognition in medical images have to be developed. Developing such methods is promising to build more accurate shape models since the mass data available nowadays can be exploited with no additional human labeling effort.



NTU Programme coordinator: Alexei Sourin (till 31 May 2018)

Programme Management Committee NTU - TU Darmstadt

Alexei Sourin (SCSE NTU) - Chair

Ng Wee Keong (SCSE NTU)

Reiner HÄhnle (TU Darmstadt)

Michael Goesele (TU Darmstadt)


Joint Admission Committee NTU - TU Darmstadt

Alexei Sourin (SCSE NTU) - Chair

Ng Wee Keong (SCSE NTU)

Reiner HÄhnle  (TU Darmstadt)

Michael Goesele (TU Darmstadt)