The whole dataset consists of three parts:

• Choreography
• Music
• Motion Capture Data

There are four files in the data/choreography directory, corresponding to four different types of dances: Cha-cha, Rumba, Tango and Waltz.

Each file consists an array of choreography objects, and each object has the following attributes:

1. name: the name of the current dance
2. dance_type: one of the four dance types that this dance belongs to.
3. dance_id: the id of this dance.
4. start_pos: the start music frame number of this dance, the FPS is 25. The first CAU starts at this frame.
5. end_pos: the end music frame number of this dance, the FPS is 25. The last CAU ends at this frame.
6. movements: the CAU sequence of this dance.
7. durations: the number of beats of each CAU in this dance.

The corresponding music of each dance. For example, if the type of this dance is Cha-cha and its dance_id is 1, its music is data/music/C/1.mp3.

data/movement_interval.csv records the motion capture clip of each CAU. The movement_tag attribute is the name of the CAU, and its motion capture clip is in the file denoted by c3d_file attribute, and the motion capture clip starts at start_frame and ends at end_frame. The c3d file is located in data/raw_c3d directory.