Conditional Fingerprint Generation
A conditional StyleGAN model generating synthetic fingerprints across varied surface backgrounds.
Python StyleGAN PyTorch Generative Models
Overview
A conditional generative model producing synthetic fingerprint images across varied surface backgrounds, built with StyleGAN and custom training loops during MSc research. The work explored whether a generator can be conditioned to produce usable biometric samples on backgrounds it was not directly trained against.
Key Features
- StyleGAN generator with custom training loops for controlled conditioning
- Background-conditioned synthesis producing 200+ fingerprints across differing surface textures
- Artefact evaluation protocol scoring each generated sample for visual defects
Results
Generated 200+ synthetic fingerprints with a 97.83% no-artefact success rate, demonstrating controlled custom-model training and benchmark evaluation.