Dataset structure

The organization of the dataset is illustrated in the left figure.
In the root folder you will find all the screenplay folders, License.txt file, and a Readme.txt file describing the contents.
For every screenplay, the relevant material is enclosed in a single folder, with sub-folders specifying the data collected for a different role and a different take.
In every sub-folder, the raw folder contains the recorded RGB-D videos in MKV format and the audio files in WAV format, while the ply folder includes the generated point cloud sequences that are stored in PLY format with a binary little endian codification.
For the ply files, a naming convention indicating the timestamp of every individual frame is followed (i.e., pointcloud-timestamp.ply).