Unsupervised Domain Adaptation with Dynamic Clustering and Contrastive Refinement for Gait Recognition

TL;DR

This work tackles unsupervised gait recognition across domain shifts by addressing noisy pseudo-labels and clothing-induced variation. It introduces GaitDCCR, combining Dynamic Clustering Parameters (DCP), Dynamic Weighted Centroids (DWC), Confidence-Based Pseudo-Label Refinement (CPR), and a Contrastive Teacher Module (CTM) within a teacher-student framework, along with targeted data augmentation and smoothed soft pseudo-labels. Ablation and extensive cross-dataset experiments (CASIA-B, OUMVLP, GREW) demonstrate robust improvements over prior unsupervised domain adaptation methods, especially under clothing variation. The results indicate that dynamic clustering and label refinement significantly enhance clustering reliability and training effectiveness, enabling strong real-world applicability for unsupervised gait recognition.

Abstract

Gait recognition is an emerging identification technology that distinguishes individuals at long distances by analyzing individual walking patterns. Traditional techniques rely heavily on large-scale labeled datasets, which incurs high costs and significant labeling challenges. Recently, researchers have explored unsupervised gait recognition with clustering-based unsupervised domain adaptation methods and achieved notable success. However, these methods directly use pseudo-label generated by clustering and neglect pseudolabel noise caused by domain differences, which affects the effect of the model training process. To mitigate these issues, we proposed a novel model called GaitDCCR, which aims to reduce the influence of noisy pseudo labels on clustering and model training. Our approach can be divided into two main stages: clustering and training stage. In the clustering stage, we propose Dynamic Cluster Parameters (DCP) and Dynamic Weight Centroids (DWC) to improve the efficiency of clustering and obtain reliable cluster centroids. In the training stage, we employ the classical teacher-student structure and propose Confidence-based Pseudo-label Refinement (CPR) and Contrastive Teacher Module (CTM) to encourage noisy samples to converge towards clusters containing their true identities. Extensive experiments on public gait datasets have demonstrated that our simple and effective method significantly enhances the performance of unsupervised gait recognition, laying the foundation for its application in the real-world. We will release the code at https://github.com/YanSun-github/GaitDCCR upon acceptance.

Figures (8)

• Figure 1: Visualization of clustering results for the previous model and our proposed model. In the figure, samples of different colors represent different subjects, and circles represent clustering results.
• Figure 2: The cluster centroid weighting strategy of previous methods and ours.
• Figure 3: The framework of our model. “DA” stands for Data Augmentation, and “EMA” represents Exponential Moving Average, which is the parameter update method of the teacher encoder.
• Figure 4: The generation process of soft pseudo-labels. “DA sample” is obtained by query sample through data augmentation, which simulates the change of clothing.
• Figure 5: The visualization of data augmentation on NM and CL conditions.