Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning Commonality, Divergence and Variety for Unsupervised Visible-Infrared Person Re-identification

Jiangming Shi, Xiangbo Yin, Yachao Zhang, Zhizhong Zhang, Yuan Xie, Yanyun Qu

TL;DR

This work proposes a Progressive Contrastive Learning with Hard and Dynamic Prototypes method for USVI-ReID that generates the hard prototype by selecting the sample with the maximum distance from the cluster center and theoretically shows that the hard prototype is used in the contrastive loss to emphasize divergence.

Abstract

Unsupervised visible-infrared person re-identification (USVI-ReID) aims to match specified people in infrared images to visible images without annotations, and vice versa. USVI-ReID is a challenging yet under-explored task. Most existing methods address the USVI-ReID using cluster-based contrastive learning, which simply employs the cluster center as a representation of a person. However, the cluster center primarily focuses on commonality, overlooking divergence and variety. To address the problem, we propose a Progressive Contrastive Learning with Hard and Dynamic Prototypes method for USVI-ReID. In brief, we generate the hard prototype by selecting the sample with the maximum distance from the cluster center. We theoretically show that the hard prototype is used in the contrastive loss to emphasize divergence. Additionally, instead of rigidly aligning query images to a specific prototype, we generate the dynamic prototype by randomly picking samples within a cluster. The dynamic prototype is used to encourage the variety. Finally, we introduce a progressive learning strategy to gradually shift the model's attention towards divergence and variety, avoiding cluster deterioration. Extensive experiments conducted on the publicly available SYSU-MM01 and RegDB datasets validate the effectiveness of the proposed method.

Learning Commonality, Divergence and Variety for Unsupervised Visible-Infrared Person Re-identification

TL;DR

This work proposes a Progressive Contrastive Learning with Hard and Dynamic Prototypes method for USVI-ReID that generates the hard prototype by selecting the sample with the maximum distance from the cluster center and theoretically shows that the hard prototype is used in the contrastive loss to emphasize divergence.

Abstract

Unsupervised visible-infrared person re-identification (USVI-ReID) aims to match specified people in infrared images to visible images without annotations, and vice versa. USVI-ReID is a challenging yet under-explored task. Most existing methods address the USVI-ReID using cluster-based contrastive learning, which simply employs the cluster center as a representation of a person. However, the cluster center primarily focuses on commonality, overlooking divergence and variety. To address the problem, we propose a Progressive Contrastive Learning with Hard and Dynamic Prototypes method for USVI-ReID. In brief, we generate the hard prototype by selecting the sample with the maximum distance from the cluster center. We theoretically show that the hard prototype is used in the contrastive loss to emphasize divergence. Additionally, instead of rigidly aligning query images to a specific prototype, we generate the dynamic prototype by randomly picking samples within a cluster. The dynamic prototype is used to encourage the variety. Finally, we introduce a progressive learning strategy to gradually shift the model's attention towards divergence and variety, avoiding cluster deterioration. Extensive experiments conducted on the publicly available SYSU-MM01 and RegDB datasets validate the effectiveness of the proposed method.
Paper Structure (17 sections, 23 equations, 3 figures, 2 tables)

This paper contains 17 sections, 23 equations, 3 figures, 2 tables.

Table of Contents

  1. Introdction
  2. Related Work
  3. Supervised Visible-Infrared Person ReID
  4. Unsupervised Single-Modality Person ReID
  5. Unsupervised Visible-Infrared Person ReID
  6. Method
  7. Problem Formulation and Overview
  8. Centroid Prototype Contrastive Learning
  9. Hard Prototype Contrastive Learning
  10. Dynamic Prototype Contrastive Learning
  11. Progressive Contrastive Learning
  12. Experiment
  13. Experiment Setting
  14. Results and Analysis
  15. Ablation Study
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Framework of our PCLHD. The framework consists of two stages: the first stage employs contrastive learning with centroid prototypes to learn well-discriminative representation, and the second stage introduces contrastive learning with hard and dynamic prototypes to further focus on divergence and variety.
  • Figure 2: (a) The effect of hyper-parameter $\lambda$ with different values. (b) The effect of hyper-parameter $k$ with different values. (c) Comparisons with ARI values of different methods.
  • Figure 3: The t-SNE visualization of 10 randomly selected identities. Different color indicates different IDs. Circle means visible features and the pentagram means infrared features.