MPI 全称 Moldflow Plastic Insight，这是由 100 多万张物理 3D 物体图像组成的数据集。图像有七个变化因素，如物体的颜色、形状、大小和位置等。该数据集可用于测试模拟环境和真实环境中的表征学习算法。 该数据集有四个子集：Real world simple shapes (mpi3d_real) 、 Realistic rendered images (mpi3d_realistic) 、 Simplistic rendered images (mpi3d_toy) 和 Complex real world shapes (mpi3d_complex)，前三个包含 1,036,800 张图像，最后一个包含 460,800 张图像。

Citation

@inproceedings{NEURIPS2019_d97d404b, author = {Gondal, Muhammad Waleed and Wuthrich, Manuel and Miladinovic, Djordje and Locatello, Francesco and Breidt, Martin and Volchkov, Valentin and Akpo, Joel and Bachem, Olivier and Sch\”{o}lkopf, Bernhard and Bauer, Stefan}, booktitle = {Advances in Neural Information Processing Systems}, editor = {H. Wallach and H. Larochelle and A. Beygelzimer and F. d\textquotesingle Alch'{e}-Buc and E. Fox and R. Garnett}, pages = {}, publisher = {Curran Associates, Inc.}, title = {On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset}, url = {https://proceedings.neurips.cc/paper/2019/file/d97d404b6119214e4a7018391195240a-Paper.pdf}, volume = {32}, year = {2019} }