Gait3D

我们构建了第一个基于大规模3D表示的步态识别数据集，名为Gait3D。它包含4,000个主题和超过25,000个序列从39摄像机在不受约束的室内场景中提取。更重要的是，它提供了从视频帧中恢复的3D SMPL模型，该模型可以提供身体形状，视点和动力学的密集3D信息。此外，它还提供了2D轮廓和关键点，可以使用多模态数据探索步态识别。基于Gait3D，我们将我们的方法与现有的步态识别方法进行了全面比较，这反映了我们框架的卓越性能以及3D表示在野外进行步态识别的潜力。

We present the first large-scale 3D representation-based gait recognition dataset named Gait3D. It contains 4,000 subjects and over 25,000 gait sequences captured by 39 cameras in unconstrained indoor scenarios. More importantly, it provides 3D SMPL models reconstructed from video frames, which can offer dense 3D information regarding body shape, viewpoint and motion dynamics. In addition, it also supplies 2D silhouettes and keypoints, which enable gait recognition exploration using multi-modal data. Based on Gait3D, we conducted comprehensive comparisons between our proposed method and existing gait recognition approaches, which demonstrates the superior performance of our framework and the potential of 3D representations for gait recognition in the wild.

搜集汇总

背景与挑战

背景概述

Gait3D是首个基于大规模3D表示的步态识别数据集，包含4,000个主题和超过25,000个序列，从39个摄像机在不受约束的室内场景中采集。它提供了3D SMPL模型、2D轮廓和关键点，支持多模态数据探索，旨在推动野外步态识别研究，相关论文于2022年CVPR发布。