Sara motion is a 3D motion dataset, named Synthetic Actors and Real Actions (SARA), for training a model to produce motion embeddings suitable for reasoning about motion similarity.

The motion sequence data for this dataset was generated by combining 18 different actors (i.e., action performing characters). The characters were rendered in a skeleton shape with Adobe Fuse software. Four action categories were selected (Combat, Adventure, Sport, and Dance) comprising a number of motion variations, where each action has a frame length of 32 or more. There are 4,428 base motions (e.g., dancing, jumping) in the SARA dataset.