[Paper Review] Biometric Quality: Review and Application to Face Recognition with FaceQnet

Javier Hernandez‐Ortega, Javier Galbally+2 more|arXiv (Cornell University)|Jun 5, 2020

Face recognition and analysis49 references58 citations

TL;DR

This paper reviews biometric quality concepts and introduces FaceQnet, a deep learning-based open-source tool for predicting face recognition accuracy from a face image, with v0 and improved v1 versions evaluated against state-of-the-art metrics.

ABSTRACT

"The output of a computerised system can only be as accurate as the information entered into it." This rather trivial statement is the basis behind one of the driving concepts in biometric recognition: biometric quality. Quality is nowadays widely regarded as the number one factor responsible for the good or bad performance of automated biometric systems. It refers to the ability of a biometric sample to be used for recognition purposes and produce consistent, accurate, and reliable results. Such a subjective term is objectively estimated by the so-called biometric quality metrics. These algorithms play nowadays a pivotal role in the correct functioning of systems, providing feedback to the users and working as invaluable audit tools. In spite of their unanimously accepted relevance, some of the most used and deployed biometric characteristics are lacking behind in the development of these methods. This is the case of face recognition. After a gentle introduction to the general topic of biometric quality and a review of past efforts in face quality metrics, in the present work, we address the need for better face quality metrics by developing FaceQnet. FaceQnet is a novel open-source face quality assessment tool, inspired and powered by deep learning technology, which assigns a scalar quality measure to facial images, as prediction of their recognition accuracy. Two versions of FaceQnet have been thoroughly evaluated both in this work and also independently by NIST, showing the soundness of the approach and its competitiveness with respect to current state-of-the-art metrics. Even though our work is presented here particularly in the framework of face biometrics, the proposed methodology for building a fully automated quality metric can be very useful and easily adapted to other artificial intelligence tasks.

Motivation & Objective

• Explain biometric quality and its importance for face recognition.
• Propose a fully automated, scalable face quality metric using deep learning.
• Develop and evaluate FaceQnet with groundtruth derived from ICAO compliance without human labelling.
• Compare FaceQnet against state-of-the-art face quality metrics and provide open-source resources.

Proposed method

• Describe a general framework for biometric quality as a predictor of recognition accuracy.
• Develop FaceQnet by transferring knowledge from deep face representations to a quality predictor.
• Create groundtruth quality scores automatically using ICAO/compliance-based labeling.
• Evaluate two FaceQnet versions (v0 and v1) against state-of-the-art metrics and independent evaluation by NIST FRVT.
• Release FaceQnet as an open-source project and provide quality labels for LFW and VGGFace2.

Experimental results

Research questions

• RQ1 How can biometric quality be quantified as a predictor of face recognition accuracy?
• RQ2 Can a deep learning-based quality metric outperform existing hand-crafted or traditional metrics for face quality?
• RQ3 Do two versions of FaceQnet (v0 and v1) show improved alignment with recognition performance?
• RQ4 Is the approach robust across multiple face databases and independent evaluations?
• RQ5 Can the methodology generalize to other AI tasks beyond face recognition?

Key findings

• FaceQnet provides a numerical quality score from 0 to 1 that correlates with recognition accuracy.
• FaceQnet v1 shows improvement over FaceQnet v0 in evaluation and provides competitive performance against state-of-the-art metrics.
• Independent evaluation by NIST FRVT corroborates the soundness and competitiveness of FaceQnet.
• FaceQnet is released as open source, enabling broader adoption and adaptation to other AI tasks.
• Quality labels for LFW and VGGFace2 are generated and made available with FaceQnet.