Template-based Multi-Domain Face Recognition
This addresses face recognition for security and surveillance applications in diverse domains, but it is incremental as it builds on existing pooling methods.
The paper tackled face recognition in challenging non-visible domains like SWIR and surveillance, where training data is scarce, by introducing a template generation algorithm called Norm Pooling that outperformed average pooling on the IJB-MDF dataset.
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.