Self-supervised random mask attention GAN in tackling pose-invariant face recognition

Download PDF
Open Webpage

Jiashu Liao, Tanaya Guha, Victor Sanchez

Published: 2025, Last Modified: 07 Nov 2025Pattern Recognit. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Introduced the “Mask Rotate” framework for PIFR using unsupervised learning, transforming PIFR into face image completion.•Eliminated the double rotation process in training by introducing the “Random Mask” method.•Introduced the “Random Mask Attention Generative Adversarial Network (RMAGAN)” for high-quality image completion using landmark-guided attention.•Utilized diffusion models for high-resolution post-processing.•The framework does not require paired training data, making it suitable for real-world applications.