Age-Invariant Face Recognition Based on Identity-Age Shared Features

Age-invariant face recognition is a challenging task due to the significant facial changes caused by aging. This paper introduces a novel approach based on identity-age shared features, leveraging multi- vision transformers for robust recognition across age variations. Age-invariant face recognition (AIFR) has gained considerable attention due to its crucial role in identity verification across varying age ranges. Traditional convolutional neural networks (CNNs) have been widely employed for AIFR; however, their limitations in capturing long-range dependencies and facial dynamics across different age groups motivate the exploration of alternative approaches. This work presents a novel AIFR framework based on multi-vision transformers (ViTs) that leverage identity-age shared features without the use of CNNs. By integrating multi-scale vision transformers, our approach captures global contextual information and inherent facial patterns that remain stable across age progression. The proposed model employs a shared feature representation strategy that unifies age-invariant characteristics while maintaining the distinctiveness of individual identities. Extensive experimentation on public datasets demonstrates that the proposed ViT-based framework achieves competitive performance, surpassing state-of-the-art CNN-based methods in both accuracy and age-invariance. Our study highlights the potential of vision transformers in addressing the challenges of AIFR without reliance on traditional convolutional architectures.