BarbieGait: An Identity-Consistent Synthetic Human Dataset with Versatile Cloth-Changing for Gait Recognition

TL;DR

BarbieGait is a synthetic dataset for gait recognition that simulates clothing changes to evaluate and improve cross-clothing gait identification methods.

Contribution

The paper introduces BarbieGait, a controllable synthetic gait dataset with clothing variation, and proposes GaitCLIF, a model for cloth-invariant gait recognition.

Findings

GaitCLIF significantly improves cross-clothing recognition accuracy.

BarbieGait enables large-scale testing of clothing-invariant gait features.

The dataset helps validate gait recognition methods under diverse clothing conditions.

Abstract

Gait recognition, as a reliable biometric technology, has seen rapid development in recent years while it faces significant challenges caused by diverse clothing styles in the real world. This paper introduces BarbieGait, a synthetic gait dataset where real-world subjects are uniquely mapped into a virtual engine to simulate extensive clothing changes while preserving their gait identity information. As a pioneering work, BarbieGait provides a controllable gait data generation method, enabling the production of large datasets to validate cross-clothing issues that are difficult to verify with real-world data. However, the diversity of clothing increases intra-class variance and makes one of the biggest challenges to learning cloth-invariant features under varying clothing conditions. Therefore, we propose GaitCLIF (Gait-oriented CLoth-Invariant Feature) as a robust baseline model for cross-clothing gait recognition. Through extensive experiments, we validate that our method significantly improves cross-clothing performance on BarbieGait and the existing popular gait benchmarks. We believe that BarbieGait, with its extensive cross-clothing gait data, will further advance the capabilities of gait recognition in cross-clothing scenarios and promote progress in related research.