Matching 3D Facial Shape to Demographic Properties by Geometric Metric Learning: A Part-Based Approach

• NCJ Number: 307358
• Journal: IEEE Transactions on Biometrics, Behavior, and Identity Science Volume: 4 Issue: 2 Dated: April 2022 Pages: 163-172
• Author(s): Soha Sadat Mahdi; Nele Nauwelaers; Philip Joris; Giorgos Bouritsas; Shunwang Gong; Susan Walsh; Mark D. Shriver; Michael Bronstein; Peter Claes
• Date Published: April 2022
• Length: 10 pages
• Grant Number(s): 2018-DU-BX-0219; 2014-DN-BX-K031
• Agencies: NIJ-Sponsored
• Publication Type: Research (Applied/Empirical)
• Award: Improving the prediction of human quantitative pigmentation traits such as eye, hair and skin color using a worldwide representation panel of US and European individuals; Ensuring standards and enhancing the field of predictive biometrics using a globally-diverse genotype-phenotype database

Annotation

This paper presents research and the development of a method for allowing a global-to-local part-based analysis of the face; the authors report on their evaluation of the capacity of the model for establishing identity from facial shape against a list of probe demographics; and they discuss the results of their efforts.

Abstract

Face recognition is a widely accepted biometric identifier, as the face contains a lot of information about the identity of a person. The goal of this study was to match the 3D face of an individual to a set of demographic properties (sex, age, body-mass index (BMI), and genomic background) that are extracted from unidentified genetic material. In this document, the authors introduce a triplet loss metric learner that compresses facial shape into a lower dimensional embedding while preserving information about the property of interest. The metric learner is trained for multiple facial segments to allow a global-to-local part-based analysis of the face. To learn directly from 3D mesh data, the authors used spiral convolutions along with a novel mesh-sampling scheme, which retains uniformly sampled points at different resolutions. The authors evaluate the capacity of the model for establishing identity from facial shape against a list of probe demographics by enrolling the embeddings for all properties into a support vector machine classifier or regressor and then combining them using a naive Bayes score fuser. Results obtained by a 10-fold cross-validation for biometric verification and identification show that part-based learning significantly improves the systems performance for both encoding with our geometric metric learner or with principal component analysis. Publisher Abstract Provided