Multi-state perception consistency constraints network for person re-identification

Person re-identification (Re-ID) remains challenging due to pose variations and scale changes across non-overlapping camera views. In this work, we propose a Multi-state Perception Consistency Constraints Network (MPCC-Net) that extracts discriminative and robust features for person Re-ID. MPCC-Net...

Full description

Saved in:
Bibliographic Details
Published inPattern analysis and applications : PAA Vol. 28; no. 1
Main Authors Zhou, Mengting, Lian, Guoyun, Ouyang, Xinyu, Du, Jingyu, Song, Qiqi, Yang, Jinfeng
Format Journal Article
LanguageEnglish
Published London Springer London 01.03.2025
Springer Nature B.V
Subjects
Body parts
Computer Science
Constraints
Original Article
Pattern Recognition
Perception
Source code
Feature enhancement
Perceptual constraint
Person re-identification
Convolutional neural network
Online AccessGet full text

Cover

More Information
Summary:Person re-identification (Re-ID) remains challenging due to pose variations and scale changes across non-overlapping camera views. In this work, we propose a Multi-state Perception Consistency Constraints Network (MPCC-Net) that extracts discriminative and robust features for person Re-ID. MPCC-Net consists of three primary components. First, a multi-state fused backbone network processes multi-scale and multi-view information. Second, perception consistency constraints enhance feature stability. Third, partition attention modules focus on different body parts to improve local discrimination. Comprehensive experiments on benchmark datasets demonstrate MPCC-Net’s competitive performance, effectively addressing pose and scale variations for accurate person Re-ID. Our source code will also be publicly available at: https://github.com/sesamecandy/MPCC-Net
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1433-7541
1433-755X
DOI:10.1007/s10044-024-01398-2