Exploring Generalized Gait Recognition: Reducing Redundancy and Noise within Indoor and Outdoor Datasets

2025-05-21

AI Summary

Generalized gait recognition suffers from poor cross-domain generalization due to severe domain shifts in viewpoint, appearance, and environment; meanwhile, joint training on multiple source datasets introduces optimization conflicts, feature redundancy, and noise interference. To address these challenges, we propose: (1) a decoupled triplet loss to mitigate gradient conflicts across heterogeneous datasets; and (2) a targeted data distillation strategy that jointly leverages feature redundancy and prediction uncertainty to prune the least informative 20% of samples. Our approach establishes a lightweight cross-dataset joint optimization framework compatible with mainstream backbone networks such as GaitBase and DeepGaitV2. Extensive experiments on four major benchmarks—CASIA-B, OU-MVLP, Gait3D, and GREW—demonstrate significant improvements in cross-domain recognition accuracy while preserving stable performance on source domains.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Computer Vision: Learning & Optimization for CV

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Generalized gait recognition, which aims to achieve robust performance across diverse domains, remains a challenging problem due to severe domain shifts in viewpoints, appearances, and environments. While mixed-dataset training is widely used to enhance generalization, it introduces new obstacles including inter-dataset optimization conflicts and redundant or noisy samples, both of which hinder effective representation learning. To address these challenges, we propose a unified framework that systematically improves cross-domain gait recognition. First, we design a disentangled triplet loss that isolates supervision signals across datasets, mitigating gradient conflicts during optimization. Second, we introduce a targeted dataset distillation strategy that filters out the least informative 20% of training samples based on feature redundancy and prediction uncertainty, enhancing data efficiency. Extensive experiments on CASIA-B, OU-MVLP, Gait3D, and GREW demonstrate that our method significantly improves cross-dataset recognition for both GaitBase and DeepGaitV2 backbones, without sacrificing source-domain accuracy. Code will be released at https://github.com/li1er3/Generalized_Gait.

Problem

• Addressing domain shifts in generalized gait recognition
• Mitigating inter-dataset optimization conflicts and noise
• Enhancing cross-domain gait recognition efficiency

Innovation

• Disentangled triplet loss isolates supervision signals
• Targeted dataset distillation filters uninformative samples
• Unified framework enhances cross-domain gait recognition