Semi-Siamese Training for Shallow Face Learning

TL;DR

This paper introduces Semi-Siamese Training (SST), a novel method designed to improve face recognition models trained on shallow datasets with limited intra-class variation, addressing issues of overfitting and feature collapse.

Contribution

The paper proposes SST, a flexible training approach using Semi-Siamese networks and a gallery queue, effective for shallow face data and compatible with existing architectures.

Findings

Significantly improves shallow face learning performance.

Enhances training stability and reduces overfitting.

Effective also for conventional deep face datasets.

Abstract

Most existing public face datasets, such as MS-Celeb-1M and VGGFace2, provide abundant information in both breadth (large number of IDs) and depth (sufficient number of samples) for training. However, in many real-world scenarios of face recognition, the training dataset is limited in depth, i.e. only two face images are available for each ID. $\textit{We define this situation as Shallow Face Learning, and find it problematic with existing training methods.}$ Unlike deep face data, the shallow face data lacks intra-class diversity. As such, it can lead to collapse of feature dimension and consequently the learned network can easily suffer from degeneration and over-fitting in the collapsed dimension. In this paper, we aim to address the problem by introducing a novel training method named Semi-Siamese Training (SST). A pair of Semi-Siamese networks constitute the forward propagation structure, and the training loss is computed with an updating gallery queue, conducting effective optimization on shallow training data. Our method is developed without extra-dependency, thus can be flexibly integrated with the existing loss functions and network architectures. Extensive experiments on various benchmarks of face recognition show the proposed method significantly improves the training, not only in shallow face learning, but also for conventional deep face data.