3D Shapes Dataset

该数据集是一个由480,000个三维形状组成的综合数据集，这些形状是通过6个基本真实变因程序生成的。数据集包含了诸如地板色调、墙壁色调、物体色调等可控因素，以及规模、形状和方向等。规模达到了480,000个形状，其任务是用以验证和演示可解释性方法。

This dataset is a comprehensive collection of 480,000 3D shapes, which are procedurally generated via six fundamental ground-truth variation factors. It includes controllable factors such as floor hue, wall hue, and object hue, as well as scale, shape, orientation and other relevant attributes. With a total of 480,000 samples, this dataset is designed for validating and demonstrating interpretability methods.