PackIt

该数据集包含了多个包装集合，每个集合都由一系列具有特定排列（包括旋转和位置）的形状组成，这些形状能够放入一个单位尺寸的盒子内。这样的设计旨在为几何规划生成具有挑战性的包装任务。该数据集是通过使用进化算法创建的，并采用了ShapeNet数据集中的形状，后者包含了大约51,300个来自55个不同类别的形状。在规模上，该数据集分为三个子集，分别用于训练、测试和验证，包含4207、4196和4185个形状。其任务是为几何规划和包装任务提供挑战。

This dataset comprises multiple packing sets, each composed of a series of shapes with specific arrangements (including rotation and placement) that can fit into a unit-sized box. This design is intended to generate challenging packing tasks for geometric planning. The dataset was created using evolutionary algorithms, and utilizes shapes from the ShapeNet dataset, which contains approximately 51,300 shapes across 55 distinct categories. In terms of scale, the dataset is split into three subsets for training, testing, and validation, with 4207, 4196, and 4185 shapes respectively. Its purpose is to provide challenging benchmarks for geometric planning and packing tasks.