Index

Deep Learning-Based Analysis of Thermographic Mold Images for Quality Prediction in Aluminum Die Casting

1. Motivation

Following the casting of aluminum die-cast parts, several tens of thousands of thermographic images of the mold, along with associated quality data (OK/NOK + defect type), are available.

Currently, the quality assessment only takes place after the complete process. An automatic, image-based early warning system could reduce scrap, optimize process parameters, and lower costs.

2. Objectives of the Work

• Proof of a statistically robust correlation between thermograms and quality labels.
• Development, training, and evaluation of a Deep Learning model for classification (good/bad) and, if applicable, multi-label defect detection.
• Determination of relevant image areas/features using Explainable AI (XAI) methods.
• Documentation of best-practice recommendations for the industrial partner.

Generalization of Self-Supervised Vision Models in Image Retrieval

This project investigates the generalization capabilities of prominent self-supervised vision models, such as DINOv2, CLIP, and MoCo, when applied to image retrieval tasks across diverse visual domains.

Overview

Self-supervised learning (SSL) models are increasingly important for creating robust visual features. However, their performance often degrades when transferring from standard natural image datasets (like ImageNet) to more specialized domains.

Our study systematically benchmarks these models on three distinct domains:

1. Natural Images (Baseline)

2. Scene-Centric Images (e.g., Places365)

3. Artistic Images (e.g., ArtPlaces)

We compare the models’ out-of-the-box generalization and analyze the impact of finetuning on their domain adaptation, providing crucial insights into the stability and robustness of the learned representations for practical retrieval applications.

Investigating the Influence of Using Different for Detecting Parkinson’s Disease

Integrating Transformer Networks with Multi-Modal Learning for Document Layout Analysis

A Cascaded Encoder–Decoder Network for CT Image Restoration

Evaluate LORA finetuning for detection guided segmentation in CT images

Improving localization of VL models in CXRs

Patient-Specific Quality Assurance of Synthetic CT for MR-Only Radiotherapy

If you are interested in the thesis project, please send your application (CV, letter of motivation, current transcript
of records) with subject ’sCT PSQA – Thesis’ to: bernd-niklas.axer@extern.uk-erlangen.de

Data-Driven Characterization and Modeling of the Radiotherapy Workflow

If you are interested in the project, please send your curriculum vitæ to: rafael.lobao@uk-erlangen.de
Having prior experience with Python, SQL, and statistics is advantageous.

Field of Study: Medical Engineering / Data Science

AI for Inflammatory Skin Pathology: Psoriasis and Eczema Classification in Whole-Slide Images

Exploring foundation models for high-resolution whole-slide image classification, with a focus on interpretable predictions through attention maps that align with medically relevant regions.