Computer Vision and Biometrics

In the area of Computer Vision and Biometrics (CVB, also specialisation for EE/EMSYS), we have Master's assignments continuously on the following subjects:

Finger vein recognition (recognition methods and development of sensors)
Face morphing detection
Image manipulation detection (e.g. deepfake detection)
Embedded AI
Heritage biometrics (human traces on historic objects and works of art)
3D face recognition
Iris recognition
Forensic biometrics
Robust vision

If you are interested in any ofthe listed below subjects, please contact the person indicated.

Below follows a list of assignments with a concrete description.

Morphing robust face recognition

FISWG AUTO: Automatic extraction and use of forensic facial features

Diving video classification with VLMs foundation models

Understanding how video is classified

Analyzing Visual and Identity-Level Differences Between Synthetic and Real Face Datasets

Heritage biometrics – 3D acquisition design for fingerprints, scratches and toolmarks on works of art

Will it fail? Generating videos of actions fails

Deriving text descriptions from visual model's predictions

Heritage Biometrics: does every painter paint himself?

Heritage biometrics – artist profiling from 3D fingerprints

Temporal Surgical Video Segmentation for Medical Dataset Curation

Surgical Phase Pattern Recognition from Video Data

Driver Behavior Understanding for Monitoring Fatigue and Anomalous Behavior in Autonomous Driving

Discovery of missing objects in predefined scene images

Pose Estimation in Handball Videos for Personalized Training

Egocentric vision for human behaviour understanding

Understanding the Effect of Image Quality and Resolution on Skin Lesion Recognition

Recognition of Crops Relevant for Food Security from Images – Dataset Compilation and Classification Framework Design

Benchmarking Medical Class-OOD (aka Open Set Recognition) Detection

OOD detection on MediCLIP

Can Normalizing Disentangled Flows Detect OOD data?

Medical OOD detection with an Exponentially Tilted Gaussian Prior for beta-Variational Autoencoders

Can you find the Gorilla in the CT scan? (2 topics: Sample and Object level)

Why can’t Vision-Language Models like CLIP find OOD data in medical images? Or can they? How?

Can Normalizing Disentangled Flows Detect OOD data?

Disentangled Flows for Face editing

Disentangled Flows for Anomaly Detection

Self-Supervised Out-of-Distribution Detection

Unsupervised Anomaly Detection Under Domain Shifts

Medical OOD Segmentation Detection

Medical Segmentation Foundation Model Generalizability

Detecting medical distribution shifts in hyperbolic spaces

Hyperbolic VAEs for better detection of out of distribution medical images

Hyperbolic Vision Language Models (VLMs) in medical imaging

Visual Prompt Tuning for Generalized Medical Foundation Models

Medical Foundation Models and Statistics for Detecting Distribution Drifts in Medical Data

AUTO: Auto encoders in face recognition

DIET: from obese data to sparse face models

Evaluation of AI-driven interpretation on microscopic images in oncology

Visual place recognition

TURBULENCE: 2D gen – 3D pose pipeline – atmospheric turbulence

Building an Artificial Nose: Development of an Automated Odour-Control Chamber

Teaching Computers to Smell: Machine Learning for Artificial Odour Recognition