Hoppa till huvudinnehåll

Calender

10

June

Master Thesis Presentation

Tid: 2025-06-10 13:15 till 14:30 Seminarium

Joakim Weiner presents his master thesis with title "Outpainting Fingerprint Images: Expanding Partial Prints with Generative Models"

Abstract:

Partial fingerprints are commonly encountered in forensic investigations but are often insufficient for accurate identification due to their limited information content. This thesis explores the use of generative deep learning models to outpaint—i.e., plausibly complete—partial fingerprint images. The work compares multiple architectures, including conditional GANs and diffusion models, focusing on their ability to generate realistic and structurally coherent fingerprint extensions. A novel dataset and evaluation pipeline are introduced, allowing both qualitative and quantitative assessment of generated images. Metrics such as Fréchet Inception Distance (FID) and fingerprint-specific structural scores are used to benchmark performance. The study also addresses challenges such as maintaining ridge continuity and avoiding unrealistic artifacts. Results indicate that diffusion-based models outperform GAN-based approaches in visual realism and structural consistency. The findings highlight the potential of generative models for forensic fingerprint enhancement and open new avenues for improving identification from incomplete biometric data.

 

Examiner:

Karl Åström, Centre for Mathematical Sciences, Lund University

 

Supervisors:

Alexandros Sopasakis, Centre for Mathematical Sciences, Lund University

Donglin Liu, Centre for Mathematical Sciences, Lund University



Om händelsen
Tid: 2025-06-10 13:15 till 14:30

Plats
MH:309A

Kontakt
alexandros.sopasakis@math.lth.se

Sidansvarig: webbansvarig@math.lu.se | 2016-06-20