Researchers have developed FVeinSyn, a novel framework for generating synthetic finger vein images to address the scarcity of large-scale public datasets in this field. The system decouples the synthesis of vascular topology and imaging appearance, using L-systems to model vein patterns and a cascaded region-aware GAN to render realistic images. FVeinSyn can generate a substantial dataset of 500,000 images, significantly improving realism and diversity compared to existing methods. Models trained with FVeinSyn-generated data have demonstrated a notable average accuracy improvement of 27.43% across eight public datasets. AI
IMPACT Enhances the development of deep learning models for finger vein recognition by providing a large, diverse synthetic dataset.
RANK_REASON The cluster describes a new synthetic data generation framework for a specific biometric modality, detailed in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
