NTURGBD-Parsing-4K

NTURGBD-Parsing-4K.RGB和深度从NTU RGB D (60/120) 均匀采样。如图5所示，我们为配对的rgb-d数据注释了24个人体部分。分区协议遵循。列车和测试装置都有1963样品。整个数据集包含3926个样本。希望通过贡献该数据集，我们可以促进人类感知和多模式学习的发展。NTURGBD-Parsing-4K. RGB and depth data are uniformly sampled from the NTU RGB+D (60/120) datasets. As shown in Figure 5, we annotated 24 human body parts for the paired RGB-D data. The partitioning protocol follows the established convention, with both the training and test splits containing 1963 samples each. The entire dataset contains a total of 3926 samples. We hope that contributing this dataset will facilitate the advancement of research on human perception and multimodal learning.

搜集汇总

背景与挑战

背景概述

NTURGBD-Parsing-4K数据集基于NTU RGB D数据集的RGB和深度数据均匀采样，标注了24个人体部位，包含总共3926个样本，均分为训练和测试集。该数据集旨在推动人类感知和多模态学习的研究，由上海人工智能实验室、南洋理工大学和商汤科技研究所于2022年发布。