Xia et al. (2015) dataset

The dataset proposed by Xia et al. (2015) contains six content classes and eight different styles for each type of motion.

BibTex: