CarGait: Cross-Attention based Re-ranking for Gait recognition

2025-03-05

AI Summary

Existing gait recognition models achieve strong performance in Top-K retrieval but suffer from low Rank-1 accuracy, primarily due to difficulty distinguishing highly similar hard negative samples. To address this, we propose an unsupervised re-ranking method based on cross-attention mechanisms—the first to apply cross-attention to gait sequence pair modeling. By enabling fine-grained temporal stripe feature interaction, our approach enhances discriminative representation learning and serves as a lightweight, plug-and-play post-processing module requiring no additional annotations. The method is model-agnostic and compatible with diverse single-stage gait models. Extensive experiments across three major benchmarks—Gait3D, GREW, and OU-MVLP—demonstrate consistent improvements in both Rank-1 and Rank-5 accuracy when integrated with seven state-of-the-art base models. Our method significantly outperforms existing re-ranking approaches, establishing new performance baselines for unsupervised gait re-ranking.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Learning Preferences or Rankings
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• Search and Retrieval-Augmented AI: Web search models and ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Gait recognition is a computer vision task that identifies individuals based on their walking patterns. Gait recognition performance is commonly evaluated by ranking a gallery of candidates and measuring the accuracy at the top Rank-$K$. Existing models are typically single-staged, i.e. searching for the probe's nearest neighbors in a gallery using a single global feature representation. Although these models typically excel at retrieving the correct identity within the top-$K$ predictions, they struggle when hard negatives appear in the top short-list, leading to relatively low performance at the highest ranks (e.g., Rank-1). In this paper, we introduce CarGait, a Cross-Attention Re-ranking method for gait recognition, that involves re-ordering the top-$K$ list leveraging the fine-grained correlations between pairs of gait sequences through cross-attention between gait strips. This re-ranking scheme can be adapted to existing single-stage models to enhance their final results. We demonstrate the capabilities of CarGait by extensive experiments on three common gait datasets, Gait3D, GREW, and OU-MVLP, and seven different gait models, showing consistent improvements in Rank-1,5 accuracy, superior results over existing re-ranking methods, and strong baselines.

Problem

• Improves Rank-1 accuracy in gait recognition
• Addresses hard negatives in top-K predictions
• Enhances existing models with cross-attention re-ranking

Innovation

• Cross-Attention Re-ranking for gait recognition
• Leverages fine-grained correlations between gait sequences
• Enhances Rank-1,5 accuracy in existing models