On Denoising Walking Videos for Gait Recognition

2025-05-24

AI Summary

To address the degradation of gait representation robustness caused by identity-irrelevant factors—such as clothing texture and color—in gait recognition, this paper proposes DenoisingGait, the first method to integrate generative diffusion models into gait video denoising. It employs human silhouette guidance to suppress background interference and introduces a geometry-constrained multi-scale feature matching mechanism—operating both intra-frame and inter-frame—to implicitly disentangle appearance noise. Subsequently, foreground pixels are encoded into a 2D directional vector field, yielding a low-noise, highly discriminative streaming gait representation: the Gait Feature Field. Extensive experiments demonstrate state-of-the-art performance on cross-domain and in-domain recognition tasks across CCPG, CASIA-B*, and SUSTech1K benchmarks. The source code is publicly available.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

To capture individual gait patterns, excluding identity-irrelevant cues in walking videos, such as clothing texture and color, remains a persistent challenge for vision-based gait recognition. Traditional silhouette- and pose-based methods, though theoretically effective at removing such distractions, often fall short of high accuracy due to their sparse and less informative inputs. Emerging end-to-end methods address this by directly denoising RGB videos using human priors. Building on this trend, we propose DenoisingGait, a novel gait denoising method. Inspired by the philosophy that"what I cannot create, I do not understand", we turn to generative diffusion models, uncovering how they partially filter out irrelevant factors for gait understanding. Additionally, we introduce a geometry-driven Feature Matching module, which, combined with background removal via human silhouettes, condenses the multi-channel diffusion features at each foreground pixel into a two-channel direction vector. Specifically, the proposed within- and cross-frame matching respectively capture the local vectorized structures of gait appearance and motion, producing a novel flow-like gait representation termed Gait Feature Field, which further reduces residual noise in diffusion features. Experiments on the CCPG, CASIA-B*, and SUSTech1K datasets demonstrate that DenoisingGait achieves a new SoTA performance in most cases for both within- and cross-domain evaluations. Code is available at https://github.com/ShiqiYu/OpenGait.

Problem

• Excluding identity-irrelevant cues in walking videos for gait recognition
• Overcoming limitations of sparse inputs in traditional silhouette- and pose-based methods
• Reducing residual noise in diffusion features for improved gait representation

Innovation

• Uses generative diffusion models for denoising
• Introduces geometry-driven Feature Matching module
• Combines background removal with human silhouettes