DArFace: Deformation Aware Robustness for Low Quality Face Recognition

2025-05-13

AI Summary

Existing face recognition systems suffer severe performance degradation on low-quality images (e.g., surveillance or long-distance captures), primarily due to significant domain shift between training data and real-world deployments—especially the inability to model local non-rigid deformations. Method: We propose a deformation-aware robust recognition framework that jointly models global geometric transformations and explicit local elastic deformations—without requiring paired high-/low-quality samples. Our approach introduces adversarial deformation augmentation and cross-deformation-view contrastive learning to bridge the domain gap. It integrates elastic deformation modeling, spatial transformation, and contrastive learning, requiring no additional annotations or synthetic paired data. Contribution/Results: Evaluated on challenging low-quality benchmarks—including TinyFace, IJB-B, and IJB-C—our method substantially outperforms state-of-the-art approaches, demonstrating that explicit modeling of local non-rigid deformations is critical for robust face recognition under severe image degradation.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

Facial recognition systems have achieved remarkable success by leveraging deep neural networks, advanced loss functions, and large-scale datasets. However, their performance often deteriorates in real-world scenarios involving low-quality facial images. Such degradations, common in surveillance footage or standoff imaging include low resolution, motion blur, and various distortions, resulting in a substantial domain gap from the high-quality data typically used during training. While existing approaches attempt to address robustness by modifying network architectures or modeling global spatial transformations, they frequently overlook local, non-rigid deformations that are inherently present in real-world settings. In this work, we introduce DArFace, a Deformation-Aware robust Face recognition framework that enhances robustness to such degradations without requiring paired high- and low-quality training samples. Our method adversarially integrates both global transformations (e.g., rotation, translation) and local elastic deformations during training to simulate realistic low-quality conditions. Moreover, we introduce a contrastive objective to enforce identity consistency across different deformed views. Extensive evaluations on low-quality benchmarks including TinyFace, IJB-B, and IJB-C demonstrate that DArFace surpasses state-of-the-art methods, with significant gains attributed to the inclusion of local deformation modeling.

Problem

• Enhancing face recognition robustness for low-quality images
• Addressing local non-rigid deformations in real-world scenarios
• Improving performance without paired high-low quality training data

Innovation

• Adversarially integrates global and local deformations
• Enforces identity consistency with contrastive objective
• Models local elastic deformations for robustness