Synthetic Facial Aging: A Comparative Study of Stable Diffusion and StyleGAN

Resumo

This study investigates the use of generative models for identity-preserving facial aging, focusing on the progression of real facial images over time. Traditional facial recognition systems often struggle with long-term age progression, as individuals undergo gradual yet significant changes in appearance.To address this challenge, we use real images of young adults from the FFHQ, BUPT-CBFace, and LFW datasets, ensuring balanced representation across ethnicities and genders. Facial aging is simulated at five target ages (30, 40, 50, 60, and 70 years) using two generative models: Stable Diffusion and StyleGAN. The outputs are evaluated in terms of identity preservation and perceptual quality.This approach enables a controlled comparison of age progression techniques. The results contribute to the advancement of AI-driven facial aging methods, with potential applications in forensic investigations, identity verification, and digital security.