ADL4D

ADL4D是由慕尼黑工业大学创建的一个包含4D日常活动的大型数据集，专注于多主体与多对象的交互。该数据集包含75个序列，总计1.1M的RGB-D帧，涵盖手部和物体姿态以及精细的手部动作标注。数据集通过自动化的多视角多手部3D姿态标注系统进行标注，适用于手部网格恢复和手部动作分割等任务。ADL4D旨在解决现有数据集在复杂交互场景下的局限性，推动机器人学习和人机交互技术的发展。

ADL4D is a large-scale 4D daily activity dataset developed by the Technical University of Munich, focusing on interactions between multiple agents and multiple objects. It comprises 75 sequences with a total of 1.1 million RGB-D frames, and includes annotations for hand and object poses as well as fine-grained hand motions. Annotated via an automated multi-view multi-hand 3D pose annotation system, this dataset is suitable for tasks including hand mesh recovery and hand motion segmentation. ADL4D aims to address the limitations of existing datasets in complex interaction scenarios, and advance the development of robotic learning and human-computer interaction technologies.