stellarnexrobotics/FoldPlanet-500

FoldPlanet-500折叠星球数据集是一个专为具身智能人形机器人训练而设计的高质量、结构化、可学习的真实泛化场景叠衣动作数据集。它包含多角度、高分辨率、时序清晰的视频、图像序列，全身31节点动作捕捉数据，以及详尽化、步骤化、可量化的自然语言指令。数据集覆盖了多种常见衣物类型的专业级折叠流程，旨在帮助模型学习人类的行为逻辑、操作方式、物体交互特征以及任务理解能力。

FoldPlanet-500 Fold Planet dataset is a high-quality, structured, and learnable real general scene clothing folding action dataset designed specifically for training embodied intelligence humanoid robots. It includes multi-angle, high-resolution, clear temporal video and image sequences, full-body 31-node motion capture data, as well as detailed, step-by-step, quantifiable natural language instructions. The dataset covers professional folding processes for a variety of common clothing types, aiming to help models learn about human behavior logic, operation methods, object interaction features, and task understanding capabilities.