This thesis tackles radiographic flaw detection in aluminium castings by pairing a new high-resolution X-ray dataset, annotated according to the American Society for Testing and Materials (ASTM) standards, with self-supervised pretraining on synthetic radiographs. Standard detectors are benchmarked to establish baselines and reveal domain challenges. To mitigate data scarcity and class imbalance, class-specific synthetic samples are generated from a few publicly available references using Stable Diffusion and are employed for multi-positive contrastive pretraining. The resulting domain-aligned backbone, surpasses ImageNet-pretrained baseline on downstream tasks on the proposed dataset.

Date: 01.09.2025 / 15:00 Place: B-116

Green Renovator is a serious game that makes city-scale energy management tangible using a real neighborhood dataset. Players act as a local authority balancing retrofit investments, energy supply, technologies, and policy under uncertainty via a random-event system. An in-game encyclopedia and optional annual quiz support reflection. Evaluated with N=55 via SUS and custom awareness/learning and engagement instruments, results show consistent good usability and positive educational effects. Thematic insights reveal an understanding of interdependencies and call for more events/policy depth. The study offers a practical, replayable design with demonstrated promise.

Date: 28.08.2025 / 10:00 Place: A-212

This thesis explores how human values shape ethical decision-making in responsible AI (RAI) development. Using Schwartz's value theory framework, the research surveyed 80 AI professionals across various stakeholder roles, including developers, designers, managers, and data scientists, to analyze their value profiles and ethical priorities. Key findings indicate that designers valued Universalism significantly more than developers, with developers scoring lower on Self-Transcendence values. The study shows that stakeholders' gender and experience influence ethical decisions, while also highlighting different ethical priorities among roles. This research underscores the important connection between stakeholder values and ethical AI development, encouraging better collaboration in RAI development.

Date: 27.08.2025 / 15:30 Place: A-212

Modern sky surveys generate large data volumes, increasing the need for automated analyses. In this work a stellar spectral dataset from five libraries was assembled and relabelled via SIMBAD. A three-branch 1-D CNN was developed to jointly predict MK temperature and luminosity classes in a single pass. Errors were weighted in a custom loss by their distance on the MK grid. Performance remained stable under circular shifts up to 3 pixels. An ensemble with a custom ResNet50 achieved macro-F1 67.6% and kappa 98.3/88.4, and statistical gains confirmed via bootstrapping. Grad-CAM indicated features of focus overlapped with known molecular bands.

Date: 28.08.2025 / 13:30 Place: B-223

This thesis introduces a Large Language Model (LLM)-driven framework for automated augmentation policy optimization in image classification. The approach leverages LLM reasoning to generate and adapt augmentation strategies without human intervention. Two modes are explore: a before-training method that iteratively refines static LLM-generated policies, and an in-training adaptive method where policies evolve based on real-time performance feedback. By integrating semantic understanding with optimization loops, the framework tailors augmentation to dataset characteristics, enhancing model robustness and generalization compared to conventional augmentation techniques.

Date: 28.08.2025 / 14:45 Place: B-223

This study addresses critical challenges in disease progression modeling, particularly longitudinal data scarcity and class imbalance in medical imaging. The thesis proposes a novel conditional diffusion framework for synthesizing realistic disease progression sequences from cross-sectional data, utilizing ulcerative colitis endoscopic images. A diffusion model is developed with two distinct ordinal class embedding strategies that enable interpolation between discrete disease severity classes. This approach transforms static medical datasets into dynamic progression models, offering solutions for enhanced clinical training and disease understanding in medical applications.

Date: 29.08.2025 / 13:30 Place: B-223

This thesis presents a framework to enhance the quality of LLM-assisted academic peer review by addressing the lack of domain-specific evaluation criteria. While general review guidelines cover broad concerns, they often miss methodological nuances. The proposed system extracts review questions from past OpenReview evaluations and aligns them with new submissions using structured metadata such as methodology, datasets, and evaluation metrics. Experiments show that the framework improves review specificity, reduces hallucinated content, and enhances interpretability—providing more explainable and relevant automated reviews compared to baseline models.

Date: 21.08.2025 / 14:00 Place: A-212

This thesis proposes a cost-driven predictive maintenance strategy for turbofan engines using the C-MAPSS dataset. The method predicts failures within multiple horizons using LSTM models, interprets predictions via SHAP analysis, and applies maintenance decisions based on dominant sensor contributions. Various maintenance and failure cost ratios are evaluated, and strategies are compared to identify the most cost-effective approach. Results provide decision-making insights for companies with different cost priorities and operational constraints.

Date: 25.08.2025 / 15:30 Place: A-212

This study explores university students’ behavioral intention to adopt generative AI tools for learning, using an extended Technology Acceptance Model (TAM). Constructs such as trust, social influence, hedonic motivation, and task-technology fit are integrated to better understand the factors affecting adoption. Partial Least Squares Structural Equation Modeling (PLS-SEM) is applied to analyze survey data collected from students. The findings aim to provide insights into students’ perceptions and guide effective integration of generative AI in educational contexts.

Date: 27.08.2025 / 10:30 Place: A-212

This study presents a mobile authentication method based on modeling the distribution of touch behavior features. Instead of using summary statistics, the approach represents user interactions as probability distributions and compares them using KL divergence. A Siamese neural network is used to learn differences between users. The method is evaluated on a public dataset, showing improvements over baseline models in terms of error rates. Results suggest that distribution-based modeling can provide useful information for continuous authentication without requiring active user input.

Date: 21.08.2025 / 10:00 Place: A-212