The dataset contains image data of periodic microstructural representative volume elements (RVE), as well as the effective heat conductivity for multiple phase contrasts. Various features and feature descriptors (explained in the related publication) are provided, as well as the computation thereof. The features were used in a machine learning regression setting (see related publication). The data is split into two sets, one with 30.000 samples containing only one inclusion type per RVE and another set of 1.500 samples containing mixed inclusions.

The repository comes with a data loader suited for flexible indexing.