Facial Recognition Leveraging Generative Adversarial Networks

TL;DR

The paper tackles data scarcity in face recognition by proposing an end-to-end GAN-based augmentation framework that jointly optimizes image generation and recognition. It introduces a residual-embedded generator to stabilize training and a FaceNet-based Inception-ResNet-V1 discriminator guided by a salient region extractor, enabling targeted perturbations. Results show substantial accuracy gains under limited data, including a 12.7% improvement on LFW and competitive performance relative to models trained on large-scale datasets, with strong generalization from AR to other benchmarks. The work highlights the potential of combining residual GANs with embedding-based discriminators for robust small-sample face recognition and points to future directions like cross-age and dynamic-face recognition.

Abstract

Face recognition performance based on deep learning heavily relies on large-scale training data, which is often difficult to acquire in practical applications. To address this challenge, this paper proposes a GAN-based data augmentation method with three key contributions: (1) a residual-embedded generator to alleviate gradient vanishing/exploding problems, (2) an Inception ResNet-V1 based FaceNet discriminator for improved adversarial training, and (3) an end-to-end framework that jointly optimizes data generation and recognition performance. Experimental results demonstrate that our approach achieves stable training dynamics and significantly improves face recognition accuracy by 12.7% on the LFW benchmark compared to baseline methods, while maintaining good generalization capability with limited training samples.

Figures (13)

• Figure 1: GAN structure
• Figure 2: FaceNet structure
• Figure 3: General framework of face recognition method based on adversarial generative network
• Figure 4: Encoder Structure
• Figure 5: Decoder Structure