Dataset Releases (CLI)

To download data from a dataset, a release first needs to be created. Several releases can be kept locally at the same time, images are shared between the releases to reduce disk usage.

/dataset_directory           # by default ~/.darwin/datasets
   /team-name
       /dataset-name
            /images          # shared between all releases
                1.png
                2.png
                3.png
            /releases
                /latest      # symlink to the latest release (v3)
                /v3
                    1.json
                    2.json
                    3.json
                /v2
                    1.json
                    2.json
                /v1
                    1.json

Generate a new release

The command to generate new release for your dataset is `darwin dataset export

$ darwin dataset export fetal_ultrasound 0.1
Dataset test successfully exported to example-team/fetal_ultrasound:0.1

🚧

Generating a release won't automatically pull the release data in your local instance of Darwin. See the Download a dataset section to learn how to do so.

List existing releases

$ darwin dataset releases example-team/my-new-dataset
NAME                                    IMAGES     CLASSES                   EXPORT_DATE
example-team/my-new-dataset:0.4             23           9     2020-05-11 14:48:25+00:00
example-team/my-new-dataset:0.2              1           8     2020-04-28 23:50:56+00:00
example-team/my-new-dataset:0.1             22           7     2020-04-17 09:57:22+00:00