Template-based Multi-Domain Face Recognition

TL;DR

This paper introduces a novel template generation algorithm called Norm Pooling for multi-domain face recognition, demonstrating its effectiveness across challenging non-visible spectrum domains and outperforming traditional methods.

Contribution

The paper presents Norm Pooling, a new template generation method that improves face recognition performance across diverse and challenging domains, especially with limited training data.

Findings

Norm Pooling outperforms average pooling in multi-domain face recognition.

The method is effective on the IJB-MDF dataset across various non-visible domains.

Template quality is crucial as domain complexity increases.

Abstract

Despite the remarkable performance of deep neural networks for face detection and recognition tasks in the visible spectrum, their performance on more challenging non-visible domains is comparatively still lacking. While significant research has been done in the fields of domain adaptation and domain generalization, in this paper we tackle scenarios in which these methods have limited applicability owing to the lack of training data from target domains. We focus on the problem of single-source (visible) and multi-target (SWIR, long-range/remote, surveillance, and body-worn) face recognition task. We show through experiments that a good template generation algorithm becomes crucial as the complexity of the target domain increases. In this context, we introduce a template generation algorithm called Norm Pooling (and a variant known as Sparse Pooling) and show that it outperforms average pooling across different domains and networks, on the IARPA JANUS Benchmark Multi-domain Face (IJB-MDF) dataset.