Integrating Heterogeneous Modalities for Comprehensive Facial Analysis: the Heteroface Database

Abstract

Heterogeneous Face Recognition (HFR) is gradually gaining importance in various domains including biometrics and cognitive studies. Classical methods are being replaced by deep learning techniques as a means to handle varied face modalities. The paper highlights the importance of HFR and introduces a face database that incorporates diversities in spectra, illumination, and formats (viz. photograph-sketch, longitudinal, 2D-3D). The role of deep learning methods has been discussed, and its advantages over the limitations of traditional methods have been emphasized. A comprehensive real-world HFR database, such as the one presented in the paper, will likely aid algorithm development, enriched by a neural networkbased curation technique that enhances diversity by excluding similar instances. The paper underscores the role of deep learning techniques in tackling the challenges in the field of HFR and presents the database as a significant contribution to advances in HFR research and application.