Relation-Based Deep Attention Network with Hybrid Memory for One-Shot Person Re-Identification.

Runxuan Si,Yuhua Tang,Shaowu Yang,Jing Zhao

doi:10.3390/s21155113

Runxuan Si, Yuhua Tang + Show 2 more

Open Access

https://doi.org/10.3390/s21155113

Copy DOI

Abstract

One-shot person Re-identification, which owns one labeled sample among numerous unlabeled data for each identity, is proposed to tackle the problem of the shortage of labeled data. Considering the scenarios without sufficient labeled data, it is very challenging to keep abreast of the performance of the supervised task in which sufficient labeled samples are available. In this paper, we propose a relation-based attention network with hybrid memory, which can make full use of the global information to pay attention to the identity features for model training with the relation-based attention network. Importantly, our specially designed network architecture effectively reduces the interference of environmental noise. Moreover, we propose a hybrid memory to train the one-shot data and unlabeled data in a unified framework, which notably contributes to the performance of person Re-identification. In particular, our designed one-shot feature update mode effectively alleviates the problem of overfitting, which is caused by the lack of supervised information during the training process. Compared with state-of-the-art unsupervised and one-shot algorithms for person Re-identification, our method achieves considerable improvements of 6.7%, 4.6%, and 11.5% on Market-1501, DukeMTMC-reID, and MSMT17 datasets, respectively, and becomes the new state-of-the-art method for one-shot person Re-identification.

Highlights

With the development of smart vision in the field of public safety and video surveillance, person reidentification (Re-ID) [1] gradually has become an attractive research focus
Our work is aimed at the case in which only a small number of data need to be labeled with low cost, and a significant improvement can be achieved for the Re-ID task
Relation-aware global attention (RGA) makes progress in the supervised person Re-ID task, considering that our clustering-based algorithm differs from the supervised method, we naturally considered carefully designing the structure of the attention module to generalize the relation-based module

Summary

Introduction

With the development of smart vision in the field of public safety and video surveillance, person reidentification (Re-ID) [1] gradually has become an attractive research focus. One is the purely unsupervised learning (USL) method, which gradually exploits pseudo labels from the dataset by clustering strategy and similarity metric without any supervised information [7,8,9]. The other is the unsupervised domain adaptation (UDA) person Re-ID [11,12,13], which fine-tunes the model on the unlabeled target dataset after pretraining model on a labeled source dataset. In this way, more information can be used for learning. Our method can be extended to other types of Re-identification tasks

Methods

Findings

Discussion

Conclusion