In order to download a Darwin dataset, you first need to generate a release. See Releases to learn more about how dataset releases can be generated.

Use the darwin dataset pull <dataset_identifier>:<release-name> command to pull a specific release

$ darwin dataset pull example-team/my-new-dataset:0.4
100%|███████████████████████████| 100/100 [00:20<00:00,  7.18it/s]
Dataset example-team/my-new-dataset:0.4 downloaded at ~/.darwin/datasets/example-team


Note that if this was the latest release for this dataset, then this release can also be referred to it via :latest for local operations.


Only datasets exported in the Darwin (JSON) format are downloadable via darwin-py.

The optional --folders flag can be used to mirror the remote folder structure locally

The optional --video-frames flag can be used to download videos as individual frames instead of a large video file.