Template Adaptation for Face Verification and Identification

TL;DR

This paper introduces a simple yet effective template adaptation method for face recognition that outperforms existing techniques on the IJB-A dataset, highlighting the importance of transfer learning and template-specific classifiers.

Contribution

The paper demonstrates that combining deep features with linear SVMs for template adaptation significantly improves face recognition performance on the IJB-A dataset, surpassing state-of-the-art methods.

Findings

Template adaptation with SVMs outperforms previous methods.

Template size and classifier fusion impact performance.

Various methods achieve similar top performance when combined with template adaptation.

Abstract

Face recognition performance evaluation has traditionally focused on one-to-one verification, popularized by the Labeled Faces in the Wild dataset for imagery and the YouTubeFaces dataset for videos. In contrast, the newly released IJB-A face recognition dataset unifies evaluation of one-to-many face identification with one-to-one face verification over templates, or sets of imagery and videos for a subject. In this paper, we study the problem of template adaptation, a form of transfer learning to the set of media in a template. Extensive performance evaluations on IJB-A show a surprising result, that perhaps the simplest method of template adaptation, combining deep convolutional network features with template specific linear SVMs, outperforms the state-of-the-art by a wide margin. We study the effects of template size, negative set construction and classifier fusion on performance, then compare template adaptation to convolutional networks with metric learning, 2D and 3D alignment. Our unexpected conclusion is that these other methods, when combined with template adaptation, all achieve nearly the same top performance on IJB-A for template-based face verification and identification.