Leveraging Large-Scale Face Datasets for Deep Periocular Recognition via Ocular Cropping

2025-10-30

AI Summary

Periocular recognition suffers from poor generalization due to limited training data (only thousands of samples). Method: This work introduces the first large-scale periocular dataset—derived from VGGFace2—comprising over 1.9 million cropped eye-region images, enabling end-to-end training of lightweight to deep CNN architectures. It integrates large-scale periocular cropping, deep feature learning, and cross-dataset evaluation to enhance robustness under unconstrained conditions. Contribution/Results: The proposed models achieve state-of-the-art performance: 9%–15% EER on VGGFace2-Pose and 1%–2% EER on UFPR-Periocular. These results demonstrate a significant breakthrough in overcoming the small-sample bottleneck and validate the efficacy and scalability of transferring large-scale face recognition data to periocular biometrics. The approach establishes a new paradigm for low-intrusion, real-world biometric authentication—particularly suitable for mobile and unconstrained environments.

Technology Category

• Computer Vision: Biometrics, Face, Gesture & Pose
• Machine Learning: Privacy
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

We focus on ocular biometrics, specifically the periocular region (the area around the eye), which offers high discrimination and minimal acquisition constraints. We evaluate three Convolutional Neural Network architectures of varying depth and complexity to assess their effectiveness for periocular recognition. The networks are trained on 1,907,572 ocular crops extracted from the large-scale VGGFace2 database. This significantly contrasts with existing works, which typically rely on small-scale periocular datasets for training having only a few thousand images. Experiments are conducted with ocular images from VGGFace2-Pose, a subset of VGGFace2 containing in-the-wild face images, and the UFPR-Periocular database, which consists of selfies captured via mobile devices with user guidance on the screen. Due to the uncontrolled conditions of VGGFace2, the Equal Error Rates (EERs) obtained with ocular crops range from 9-15%, noticeably higher than the 3-6% EERs achieved using full-face images. In contrast, UFPR-Periocular yields significantly better performance (EERs of 1-2%), thanks to higher image quality and more consistent acquisition protocols. To the best of our knowledge, these are the lowest reported EERs on the UFPR dataset to date.

Problem

• Developing deep periocular recognition using large-scale face datasets
• Evaluating CNN architectures for ocular biometrics performance
• Addressing performance gaps between controlled and uncontrolled acquisition conditions

Innovation

• Uses large-scale VGGFace2 dataset for training
• Evaluates three CNN architectures for periocular recognition
• Extracts ocular crops from face images for analysis