Gait Recognition in the Wild: A Large-scale Benchmark and NAS-based Baseline

TL;DR

This paper introduces GREW, the first large-scale gait recognition benchmark in the wild, compiled from open-system video streams across hundreds of cameras with rich annotations including identities, sequences, boxes, and human attributes. It also presents SPOSGait, a Single Path One-Shot NAS-based baseline that automatically discovers effective 3D-convolutional architectures for gait recognition and achieves state-of-the-art results across multiple benchmarks when retrained on related datasets. Through extensive experiments, the authors analyze the impact of data scale, distractor galleries, and attribute variations on recognition performance, demonstrating the dataset’s realism and the NAS-based model’s robustness and adaptability. The work highlights practical implications for real-world gait systems, including the importance of large-scale unlabeled data, robust pre-processing, and careful augmentation, while addressing privacy and bias considerations inherent to large biometric datasets.

Abstract

Gait benchmarks empower the research community to train and evaluate high-performance gait recognition systems. Even though growing efforts have been devoted to cross-view recognition, academia is restricted by current existing databases captured in the controlled environment. In this paper, we contribute a new benchmark and strong baseline for Gait REcognition in the Wild (GREW). The GREW dataset is constructed from natural videos, which contain hundreds of cameras and thousands of hours of streams in open systems. With tremendous manual annotations, the GREW consists of 26K identities and 128K sequences with rich attributes for unconstrained gait recognition. Moreover, we add a distractor set of over 233K sequences, making it more suitable for real-world applications. Compared with prevailing predefined cross-view datasets, the GREW has diverse and practical view variations, as well as more naturally challenging factors. To the best of our knowledge, this is the first large-scale dataset for gait recognition in the wild. Equipped with this benchmark, we dissect the unconstrained gait recognition problem, where representative appearance-based and model-based methods are explored. The proposed GREW benchmark proves to be essential for both training and evaluating gait recognizers in unconstrained scenarios. In addition, we propose the Single Path One-Shot neural architecture search with uniform sampling for Gait recognition, named SPOSGait, which is the first NAS-based gait recognition model. In experiments, SPOSGait achieves state-of-the-art performance on the CASIA-B, OU-MVLP, Gait3D, and GREW benchmarks, outperforming existing approaches by a large margin. The code will be released at https://github.com/XiandaGuo/SPOSGait.

Figures (13)

• Figure 1: Examples comparison for CASIA-B CASIA-B, OU-MVLP OU-MVLP and the proposed GREW. The first two are captured under constrained environments, while the GREW is constructed in the wild. Since OU-MVLP OU-MVLP does not release RGB data, visualization results from its original paper are adopted. Faces are masked in the GREW for privacy concerns.
• Figure 2: Identities examples of the GREW dataset. The first two rows show 2 subjects with various challenges. The last row shows a subject from the distractor set. Faces are masked to protect privacy.
• Figure 3: Examples of silhouette, GEI, 2D and 3D human pose and optical flow from the GREW dataset.
• Figure 4: Age group, gender, carrying and dressing attributes in the GREW. In (c), upper body dressing styles contain long-sleeve, short-sleeve, and sleeveless, while lower body includes long-trousers, shorts, and skirts.
• Figure 5: The pipeline of gait recognition in the wild, consisting of pre-processing and recognition steps. The pre-processing part detects humans from raw sequences and provides silhouette or pose information. Given a certain probe, the recognition part performs 1:N searching from the gallery.