Index

Diffusion Model-Based 3D CT Reconstruction for Arbitrary Trajectories

This master’s thesis will develop a novel approach using diffusion models for 3D CT reconstruction. The process will begin with denoising a highly blurred and noisy initial reconstruction from FBP or SIRT, aiming to enhance the quality of reconstructions obtained from a limited number of projection data on arbitrary trajectories. This research will focus on optimizing the denoising process to improve the output, advancing CT imaging capabilities with limited input data.

Distillation Knowledge of Large Language Models for Automotive HMI Applications

Automatic Speech Recognition at Phoneme and Word-Level To Analyze Parkinson’s Disease

Speech-Based Classification of Parkinson’s Disease Under Acoustic Variability

Evaluate Simpleshot: Simple implementation of few-shot classification on CXRs

Evaluate simpleshot – a simple implementation of few -shot classification on various findings on CXRs.

This report demonstrates that given sufficient pretraining data, we can achieve comparable classification results for complex findings such as pneumothorax , etc with less than 10 images.

Android App Respiratory Diseases

If you are interested, please send an email with your transcripts to paula.andrea.perez@fau.de with the subject: RepApp BA Thesis LME.

The students are expected to have proficiency in Android programming.

Text Embedddings in Pathological Speech

Gen AI: Speech Emotion Recognition

Apply Generative AI strategies in the field of Speech Emotion Recognition

Pre-requirements:

Pattern Analysis (Mandatory)
Deep Learning (Mandatory)
Advanced Depp Learning (Optional)
Speech and Language Understanding (Optional)
Seminar on pathological speech (Optional)

Please send your grades to paula.andrea.perez@fau.de

Universal Image Artifact Reduction via Heterogeneous Mixture of Experts

Abstract:
This master thesis proposes a novel unified framework for addressing various types of image artifacts through a heterogeneous Mixture of Experts (MoE) architecture. Unlike traditional approaches that tackle specific artifacts individually, our model leverages specialized expert networks, each designed to handle distinct degradation patterns, while maintaining the efficiency. The heterogeneous nature of the experts allows for optimal handling of diverse artifact types, from compression artifacts to motion blur, within a single unified model.