Exponential Discriminative Metric Embedding in Deep Learning

TL;DR

This paper introduces IE loss, a novel deep metric learning method that improves class separation in feature space, leading to better performance in object and face recognition tasks.

Contribution

The paper proposes the IE loss function, which enhances intra-class compactness and inter-class separability in deep metric learning, outperforming existing methods.

Findings

IE loss outperforms other DML methods on multiple datasets.

Enhanced intra-class compactness and inter-class separability achieved.

Approaches state-of-the-art results in object and face recognition.

Abstract

With the remarkable success achieved by the Convolutional Neural Networks (CNNs) in object recognition recently, deep learning is being widely used in the computer vision community. Deep Metric Learning (DML), integrating deep learning with conventional metric learning, has set new records in many fields, especially in classification task. In this paper, we propose a replicable DML method, called Include and Exclude (IE) loss, to force the distance between a sample and its designated class center away from the mean distance of this sample to other class centers with a large margin in the exponential feature projection space. With the supervision of IE loss, we can train CNNs to enhance the intra-class compactness and inter-class separability, leading to great improvements on several public datasets ranging from object recognition to face verification. We conduct a comparative study of our algorithm with several typical DML methods on three kinds of networks with different capacity. Extensive experiments on three object recognition datasets and two face recognition datasets demonstrate that IE loss is always superior to other mainstream DML methods and approach the state-of-the-art results.