Related Experiment Video
Updated: Jan 15, 2026
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
1.2K
PFIG-Palm: Controllable Palmprint Generation via Pixel and Feature Identity Guidance.
Summary
This study introduces a new framework for generating realistic palmprints using Bézier curves, improving palmprint recognition accuracy by over 18% with synthetic data. The Pixel and Feature Identity Guidance (PFIG) method ensures precise identity control in synthesized palmprints.
Area of Science:
- Biometrics
- Computer Vision
- Machine Learning
Background:
- Palmprint recognition is vital for secure authentication but limited by small datasets.
- Existing synthetic palmprint methods struggle with precise identity control.
- Bézier curves offer a generative approach but lack direct identity mapping.
Purpose of the Study:
- To develop a novel framework for synthesizing realistic palmprints with controllable identities.
- To overcome the challenge of precise identity mapping in curve-driven synthetic palmprint generation.
- To enhance palmprint recognition accuracy using synthetic data.
Main Methods:
- Proposed the Pixel and Feature Identity Guidance (PFIG) framework.
- Introduced an ID Injection (IDI) module for pseudo-paired data synthesis.
- Implemented cross-domain ID consistency losses at pixel and feature levels.
Main Results:
- PFIG framework successfully synthesizes realistic palmprints with strictly governed identities.
- Achieved over 18% improvement in recognition accuracy (TAR@1e-6) using 80,000 synthetic palmprints for pre-training.
- Outperformed existing synthetic methods when trained exclusively on synthetic data.
Conclusions:
- The proposed ID-guided synthesis approach enables controllable and realistic palmprint generation.
- PFIG framework significantly enhances palmprint recognition performance, especially with limited real data.
- This method offers a viable solution for augmenting datasets in biometrics research.
