Nonlinear Local Metric Learning for Person Re-identification

TL;DR

This paper introduces a nonlinear local metric learning method that combines local metric learning and deep neural networks to enhance person re-identification accuracy across various challenging datasets.

Contribution

The paper proposes a novel NLML approach that leverages local metric learning and deep neural networks to learn multiple nonlinear transformations for improved person re-identification.

Findings

Achieves state-of-the-art results on VIPeR, GRID, and CUHK 01 datasets.

Effectively exploits discriminative information through margin enforcement.

Outperforms existing methods in person re-identification accuracy.

Abstract

Person re-identification aims at matching pedestrians observed from non-overlapping camera views. Feature descriptor and metric learning are two significant problems in person re-identification. A discriminative metric learning method should be capable of exploiting complex nonlinear transformations due to the large variations in feature space. In this paper, we propose a nonlinear local metric learning (NLML) method to improve the state-of-the-art performance of person re-identification on public datasets. Motivated by the fact that local metric learning has been introduced to handle the data which varies locally and deep neural network has presented outstanding capability in exploiting the nonlinearity of samples, we utilize the merits of both local metric learning and deep neural network to learn multiple sets of nonlinear transformations. By enforcing a margin between the distances of positive pedestrian image pairs and distances of negative pairs in the transformed feature subspace, discriminative information can be effectively exploited in the developed neural networks. Our experiments show that the proposed NLML method achieves the state-of-the-art results on the widely used VIPeR, GRID, and CUHK 01 datasets.