transform : extract keypoints from files prepare : build a train / validation dataset of keypoints files, should be run only once display : you can check if open is working train test infer the transform could be included in the pipeline