Hybrid Object Dataset

该数据集包含了一系列旨在测试富含接触操作的情景，其中包括来自YCB数据集的体积较大物品以及通过TRELLIS生成的脆弱物体，后者能够从单视角图像构建出精细的物体网格。此外，该数据集还包括了在杂乱环境中评估机器人轨迹的一系列指标，如到达目标、路径成本、接触时长和位移等。规模上，它包含了20个模拟场景和10个真实世界场景。该数据集的任务是进行富含接触操作的动态规划。

This dataset comprises a series of scenarios designed to test contact-rich manipulation tasks. It includes larger-sized objects sourced from the YCB dataset, as well as fragile objects generated via TRELLIS—a method capable of constructing high-fidelity object meshes from single-view images. Additionally, the dataset provides a set of metrics for evaluating robot trajectories in cluttered environments, including target reaching, path cost, contact duration, and displacement. In terms of scale, the dataset contains 20 simulated scenarios and 10 real-world scenarios. The core task of this dataset is dynamic planning for contact-rich manipulation operations.