The 3D-video Point Cloud Musicians Dataset

DESCRIPTION The 3D-video Point Cloud Musicians Dataset provides 3D-video point clouds of several performers playing five different instruments: cello, doublebass, guitar, saxophone, and violin. The full 3D-video recordings span 1 hour of duration with an average of 12 different performers for each instrument. All musicians point clouds are centered in the origin and the axes have the following meaning: the body face direction is the z-axis, stature direction is the y-axis, and the side direction is the x-axis. The provided musician point clouds are down-sampled from their full resolution by using a voxel size of 0.01. RECORDINGS Recordings were conducted using a single Azure Kinect DK placed one meter above the floor and capturing a frontal view of the musician at a distance of two meters. The Azure Kinect DK depth camera was capturing a 75°x65° field of view with a 640x576 resolution while the Azure Kinect DK color camera was capturing with a 1920x1080 resolution. Both cameras were recording at 15 fps and Open3D library was then used to align depth and color streams and generate a point cloud for each frame. Each frame is saved as .ply format file. NAMING CONVENTION The folders in the dataset follow this structure: Instrument -> Person. The captured frames are named according to their temporal position using five numerical characters, i.e. %5d.ply