Datumaro is a Python library and CLI tool that helps simplify and enhance the dataset preparation process for model training. Developed by Intel and CVAT, Datumaro can read and write datasets in various formats, transform and merge datasets, manage dataset versions, validate annotations, and run models on the resulting datasets. It supports images, video, and 3d point clouds. Datumaro can be used independently or together with CVAT.
You can find more information on our YouTube channel