Finding a Thesis Topic

Students who are interested in writing a bachelor’s or master’s thesis should begin thinking about possible topics (cf. hot topics for thesis projects on this page) or propose their own (cf. introduction to IML). Good research questions often have their origins in scientific papers around the research topics of the IML lab. Be on the look out for new data sources that might help provide new insights into a special IML research topic.

Your Advisor and Your Committee

In order to write a bachelor’s or master’s thesis you must find an member of the IML lab who is willing to be your thesis advisor. You propose your thesis topic together with your advisor to Prof. Sonntag as the first reviewer in your committee. 

How Long Should it Be? How Long Does it Take?

A bachelor’s thesis is generally 30-60 pages, not including the bibliography. A master’s thesis is generally 60-80 pages, not including the bibliography. However, the length will vary according to the topic and the method of analysis, so the appropriate length will be determined by you, your advisor, and your committee.  Students who write a master’s thesis generally do so over two semesters, bachelor’s one semester.

Procedure and Formal Requirements

You must maintain continuous enrollment Oldenburg University or at Saarland University while working on the bachelor’s or master’s thesis. If you are planning to conduct interviews, surveys or do other research involving human subjects, you must obtain prior approval from DFKI. 

Hot Topics for Thesis Projects

Sketches can be recorded using a digital pen on paper or styluses on tablets and smartphones. Digital pen features describe various geometrical, spacial, temporal and other properties of pen input. The goal of this thesis topic is to create an interactive pen-based user interface using the digital pen features for explainable ML-based sketch recognition. An example video of such a system is found at

Contact: Alexander Prange

In this thesis state of the art handwriting recognition in combination with machine learning approaches will be used for automated spell checking, where users can interactively train both the gesture recognition and the spell checking models. The spell checking aspect can be combined with handwriting generation through deep learning networks to provide a fully interactive correction platform. See also and

Contact: Alexander Prange

In this thesis, you will implement a system that adapts the visualization of a document based on real-time relevance feedback. You will use/implement a machine learning model that predicts text relevance based on the users eye movements to, e.g., highlight relevant or hide irrelevant information from the user.

Anna Maria Feit, Lukas Vordemann, Seonwook Park, Caterina Berube, and Otmar Hilliges. 2020. Detecting relevance during decision-making from eye movements for UI adaptation. In Eye Tracking Research and Applications Symposium (ETRA), 1–11.

Nilavra Bhattacharya, Somnath Rakshit, and Jacek Gwizdka. 2020. Towards Real-time Webpage Relevance Prediction Using Convex Hull Based Eye-tracking Features. In ACM Symposium on Eye Tracking Research and Applications, 10.

Contact: Michael Barz

You will implement novel modern approaches in computer vision such as Transfer Learning, Graph Neural Network, or Semi-Supervised Learning to solve important medical decision problems like Breast cancer detection, Chest-(X-Ray/CT) abnormalities diagnosis, or related medical domains. The target is to achieve state-of-the-art performance and the proposed method could be explainable to end users to improve the system’s reliability.

Nguyen, Duy MH, et al. “An Attention Mechanism using Multiple Knowledge Sources for COVID-19 Detection from CT Images.”,  AAAI 2021, Workshop: Trustworthy AI for Healthcare. 

Soberanis-Mukul, Roger D., Nassir Navab, and Shadi Albarqouni. “An Uncertainty-Driven GCN Refinement Strategy for Organ Segmentation.” arXiv preprint arXiv:2012.03352 (2020).

Contact: Duy Nguyen

In this topic, we will investigate important theoretical machine learning problems that have high impacts on several medical applications. It includes but is not limited to optimization formulation to incorporate efficient user’s feedback to boost the performance of trained models besides available training data (active learning), investigate benefits of transfer learning strategies when dealing with scarce data issues in medical problems, or training algorithms to adapt with highly imbalanced data distribution.

Wilder, Bryan, Eric Horvitz, and Ece Kamar. “Learning to complement humans.” arXiv preprint arXiv:2005.00582 (2020).

De, Abir, et al. “Classification Under Human Assistance.” AAAI (2021).

Yao, Huaxiu, et al. “Hierarchically structured meta-learning.” International Conference on Machine Learning. PMLR, 2019.

Contact: Duy Nguyen

Robust fixation detection is essential for analysis of mobile eye tracking data in research and for gaze-based interaction. However, most fixation detection algorithms are optimized for remote eye tracking which makes them less robust in mobile settings with users moving around. A recent approach for mobile settings addresses this issue by incorporating visual similarity of gaze targets to identify fixation events more robustly [1]. But, it was not evaluated in real-time interactive settings. Your task in this thesis will be to implement the proposed algorithm (and optionally an improved version) and to systematically evaluate it in mobile and natural interaction settings.

[1] Julian Steil, Michael Xuelin Huang, and Andreas Bulling. 2018. Fixation detection for head-mounted eye tracking based on visual similarity of gaze targets. In Eye Tracking Research and Applications Symposium (ETRA), 1–9. 

Contact: Michael Barz

Inside this topic, we would investigate the image caption generation problem under the low resource setting, i.e., unavailability of large amount of paired data for training the model. For solving this challenge, we will explore and study the impact of different techniques from the areas of transfer learning, active learning. The goal of the thesis would be to develop an image caption system that achieves state-of-the-art performance under this low resource constraint.

Contact: Rajarshi Biswas

Under this topic, we would investigate interactive machine learning techniques for jointly clustering unpaired images and text. The goal would be to develop a deep learning algorithm that incorporates human-in-the-loop interaction to guide the joint learning of concepts from images and text for achieving clusters/groups having common concepts.

Contact: Rajarshi Biswas

The goal of this thesis is to develop an intelligent personnel planning system for flexible shift scheduling in nursing, which above all takes into account the interests of the employees. The shortage of qualified nursing personnel is a major topic that shapes public debate and political agenda around the globe. A possible focus for the thesis would be the comparison of different state-of-the-art planning techniques, such as SAT solvers and Deep Learning approaches.

Contact: Alexander Prange

The goal of this thesis is to develop an intelligent personnel planning system for flexible shift scheduling in nursing, which above all takes into account the interests of the employees. The shortage of qualified nursing personnel is a major topic that shapes public debate and political agenda around the globe. A possible focus for the thesis would be the development and evaluation of a chatbot which supports employees which use the digital personnel planning system.

Contact: Alexander Prange

There are various consumer-grade hardware devices available that capture handwriting and sketching digitally, such as digital pens, digitizer tablets, smartphones. In previous work we have shown that an automatic analysis of handwriting and sketches can provide vital insights about a subject’s cognitive state. As part of this thesis, we want to investigate unobtrusive and transparent estimation of cognitive state during routine everyday tasks like note taking or sketching. This could be used for various purposes, including performance evaluation or mood estimation.

Contact: Alexander Prange

As part of our research projects in the medical domain it is often necessary to annotate medical images with additional information. Deep learning systems for automatic segmentation and classification can help in determining regions of interest in these pictures. The focus of this thesis would be on designing and implementing an interactive annotation system that employs the output of deep learning networks to improve the annotation process for the user. It is possible to include multimodal interaction, such as speech, eye-tracking or digital pens to further enhance user experience.

Contact: Alexander Prange