Learning Diverse Features with Part-Level Resolution for Person Re-Identification

TL;DR

This paper introduces PLR-OSNet, a lightweight network that combines global and local features with part-level resolution, achieving state-of-the-art person re-identification performance with a simple, efficient architecture.

Contribution

The paper proposes a novel lightweight network architecture, PLR-OSNet, that effectively combines global and local features for person re-identification, emphasizing part-level feature resolution with a single identity prediction.

Findings

Achieves state-of-the-art results on Market1501, DukeMTMC-reID, and CUHK03 datasets.

Demonstrates that a simple, lightweight architecture can outperform more complex models.

Validates the effectiveness of part-level feature resolution in person re-identification.

Abstract

Learning diverse features is key to the success of person re-identification. Various part-based methods have been extensively proposed for learning local representations, which, however, are still inferior to the best-performing methods for person re-identification. This paper proposes to construct a strong lightweight network architecture, termed PLR-OSNet, based on the idea of Part-Level feature Resolution over the Omni-Scale Network (OSNet) for achieving feature diversity. The proposed PLR-OSNet has two branches, one branch for global feature representation and the other branch for local feature representation. The local branch employs a uniform partition strategy for part-level feature resolution but produces only a single identity-prediction loss, which is in sharp contrast to the existing part-based methods. Empirical evidence demonstrates that the proposed PLR-OSNet achieves state-of-the-art performance on popular person Re-ID datasets, including Market1501, DukeMTMC-reID and CUHK03, despite its small model size.