Teapot dataset

The dataset used in the paper is a wide domain image dataset, and the authors propose a weakly semi-supervised method for disentangling using both labeled and unlabeled data.

BibTex: