Open-Set Vein Biometric Recognition with Deep Metric Learning

2026-04-16 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Expert, quick

Summary

A new vein biometric recognition framework addresses the limitations of traditional closed-set classification by employing Deep Metric Learning (DML) for open-set scenarios. This approach learns discriminative L2-normalised embeddings and uses prototype-based matching with a calibrated similarity threshold to differentiate between enrolled users and unknown impostors. The framework was evaluated using a strict subject-disjoint protocol across four datasets: MMCBNU 6000, UTFVP, FYO, and a dorsal hand-vein dataset. On the large-scale MMCBNU 6000 benchmark, the ResNet50-CBAM model achieved an Open-Set Classification Rate (OSCR) of 0.9945, an AUROC of 0.9974, and an Equal Error Rate (EER) of 1.57%, alongside 99.6% Rank-1 identification accuracy. Cross-dataset experiments confirmed robust handling of large-scale data but noted sensitivity to domain shifts in low-data regimes.

Key takeaway

For Research Scientists developing biometric systems, this DML-based open-set approach offers a robust alternative to traditional closed-set methods, enabling adaptive enrollment of new users without retraining. You should consider implementing triplet-based objectives with a 1-NN classifier to achieve an optimal balance of accuracy and efficiency, especially for real-time deployment on standard hardware. Be mindful of performance degradation due to domain shifts in low-data environments.

Key insights

Deep Metric Learning enables scalable open-set vein biometric recognition, overcoming closed-set limitations.

Principles

• Open-set recognition improves scalability.
• DML learns discriminative embeddings.
• Prototype matching uses similarity thresholds.

Method

The method learns L2-normalised embeddings, then applies prototype-based matching with a calibrated similarity threshold to distinguish enrolled users from unseen impostors, evaluated under a subject-disjoint protocol.

In practice

• Use triplet-based objectives for DML.
• Combine with 1-NN for efficiency.
• Deploy on commodity hardware.

Topics

• Open-Set Recognition
• Vein Biometrics
• Deep Metric Learning
• L2-Normalised Embeddings
• Prototype-Based Matching

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• ConvNeXt-FD: A Fractal-Based Deep Model for Robust Biomedical Image Segmentation — ConvNeXt-FD enhances biomedical image segmentation by combining ConvNeXt with fractal-inspired boundary regularization for superior shape fidelity.
• KRONE: Scalable LLM-Augmented Log Anomaly Detection via Hierarchical Abstraction — Krone leverages hierarchical log structures and hybrid LLM-pattern matching for efficient, accurate, and interpretable anomaly detection.
• Efficient Data Selection for Multimodal Models via Incremental Optimization Utility — One-Step-Train (OST) optimizes LMM data selection by estimating marginal utility via a lightweight proxy, improving efficiency and performance.
• GOTabPFN: From Feature Ordering to Compact Tokenization for Tabular Foundation Models on High-Dimensional Data — GOTabPFN enhances small tabular foundation models for High-Dimensional, Low-Sample Size data via graph-guided feature ordering and neuro-inspired subunit compression.
• A Two-Stage, Object-Centric Deep Learning Framework for Robust Exam Cheating Detection — A two-stage, object-centric deep learning framework significantly improves exam cheating detection accuracy and scalability.
• Learning the Geometry of Data: A Mathematical Review of Shape Space Analysis — Shape space analysis offers a unified geometric framework to extract patterns from complex data, outperforming traditional ML on subtle variations and sparse labels.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.