KITchen

KITchen是一个专为厨房环境设计的6D物体姿态估计的大型真实世界数据集，由卡尔斯鲁厄理工学院的高性能人形技术实验室创建。该数据集包含约205,000个真实世界的RGBD图像，涵盖111种厨房物品，这些物品分布在两个不同的厨房环境中，通过一个人形机器人的自我中心视角捕捉。数据集的创建过程中，采用了一种半自动标注流程，以减少人工标注的工作量，生成了2D物体标签、2D物体分割掩码和6D物体姿态。KITchen数据集主要用于解决机器人移动操作任务中的6D姿态估计问题，特别是在厨房环境中，通过单一的机器人自我中心视野进行物体姿态估计。

KITchen is a large-scale real-world dataset dedicated to 6D object pose estimation in kitchen environments, developed by the High-Performance Humanoid Technology Laboratory at Karlsruhe Institute of Technology. It contains approximately 205,000 real-world RGBD images covering 111 kitchen items, distributed across two distinct kitchen environments and captured from the egocentric perspective of a humanoid robot. A semi-automatic annotation pipeline was adopted during dataset creation to reduce manual annotation workload, generating 2D object labels, 2D object segmentation masks, and 6D object poses. The KITchen dataset is primarily designed to solve 6D pose estimation problems in robotic mobile manipulation tasks, specifically for object pose estimation using a single robot egocentric view in kitchen environments.