A Multi-Task Framework Using Mamba for Identity, Age, and Gender Classification from Hand Images

Biometric authentication is crucial for secure access, surpassing traditional password-based methods vulnerable to breaches. Non-intrusive techniques like hand-based biometrics offer unique advantages, using physiological traits to identify individuals and predict soft biometrics such as age and gender. However, most current systems are designed with a single-purpose focus. In this field, Convolutional Neural Networks (CNNs) are typically utilized but struggle to capture long-range dependencies in images. Transformers, while more effective at handling such relationships, come with high computational costs. To overcome these challenges, this study introduces a novel multi-task learning framework that predicts identity, age, and gender simultaneously. The framework integrates the efficient long-range dependency modeling of Mamba, utilizing Visual State-Space Models (VSS) to capture both local and global patterns with reduced computational complexity. Experiments on the 11k Hands dataset demonstrate superior or competitive performance compared to existing methods.<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup><sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>Our code and pretrained models are available at: https://github.com/Avir-AI/handimage_mamba