A collection represents either a group of related labels or a group of related source imagery for a given time period and geographic area. All collections in the Radiant MLHub API are valid STAC Collections. For instance, the umd_mali_crop_type_source collection catalogs the source imagery associated with the 2019 Mali CropType dataset, while the umd_mali_crop_type_labels collection catalogs the land cover labels associated with this imagery. These collections are considered part of a single umd_mali_crop_type dataset (see the Datasets documentation for details on working with datasets).


The Radiant MLHub web application provides an overview of all the datasets and collections available through the Radiant MLHub API.


Collections are grouped into Datasets. See also the Datasets guide for more information about finding and downloading Datasets.

To list and fetch collections, the Collection class is the recommended approach, but there are also low-level client methods from radiant_mlhub.client. Both methods are described below.

Discovering Collections

You can discover collections using the Collection.list method. This method returns a list of Collection instances.

>>> from radiant_mlhub import Collection
>>> collections = Collection.list()
>>> first_collection = collections[0]
>>> print(first_collection)
ref_landcovernet_sa_v1_source_landsat_8: LandCoverNet South America Landsat 8 Source Imagery

Low-level client

The Radiant MLHub /collections endpoint returns a list of objects describing the available collections. You can use the low-level list_collections() function to work with these responses as native Python data types (list and dict). This function returns a list of JSON-like dictionaries representing STAC Collections.

>>> from radiant_mlhub.client import list_collections
>>> from pprint import pprint
>>> collections = list_collections()
>>> first_collection = collections[0]
>>> pprint(first_collection)
{'description': 'LandCoverNet South America Landsat 8 Source Imagery',
 'id': 'ref_landcovernet_sa_v1_source_landsat_8',

Fetching Collection Metadata

You can fetch a collection from the Radiant MLHub API based on the collection ID using the Collection.fetch method. This is the recommended way of fetching a collection. This method returns a Collection instance. Fetching returns the metadata but does not download assets.

>>> collection = Collection.fetch('ref_african_crops_kenya_01_labels')
>>> print(collection)
ref_african_crops_kenya_01_labels: African Crops Kenya

For more information on a collection, you can browse to the MLHub page for the related dataset, for example:

>>> print(collection.registry_url)

Browse to

Low-level client

The Radiant MLHub /collections/{id} endpoint returns an object representing a single collection’s metadata. You can use the low-level get_collection() function to work with this response as a dict.

>>> from radiant_mlhub.client import get_collection
>>> collection = get_collection('ref_african_crops_kenya_01_labels')
>>> pprint(collection)
{'description': 'African Crops Kenya',
'extent': {'spatial': {'bbox': [[34.18191992149459,
            'temporal': {'interval': [['2018-04-10T00:00:00Z',
'id': 'ref_african_crops_kenya_01_labels',

Downloading a Collection


Not all collections have downloadable archives (depending on size). Consider instead using the dataset downloader functionality. The Datasets guide has more examples and the API reference is available as well.

You can download a collection archive using the method. This is the recommended way of downloading a collection archive.


To check the existence, and size of the download archive without actually downloading it, you can use the Collection.archive_size property, which returns a size in bytes.

>>> collection = Collection.fetch('sn1_AOI_1_RIO')
>>> collection.archive_size
>>> archive_path ='~/Downloads')
28%|██▊       | 985.0/3496.9 [00:35<00:51, 48.31M/s]
>>> archive_path

If a file of the same name already exists, these methods will check whether the downloaded file is complete by comparing its size against the size of the remote file. If they are the same size, the download is skipped, otherwise the download will be resumed from the point where it stopped. You can control this behavior using the if_exists argument. Setting this to "skip" will skip the download for existing files without checking for completeness (a bit faster since it doesn’t require a network request), and setting this to "overwrite" will overwrite any existing file.

Collection archives are gzipped tarballs. You can read more about the structure of these archives in this Medium post.

Low-level client

The Radiant MLHub /archive/{archive_id} endpoint allows you to download an archive of all assets associated with a given collection. You can use the low-level download_collection_archive() function to download the archive to your local file system.

>>> from radiant_mlhub.client import download_collection_archive
>>> archive_path = download_collection_archive('sn1_AOI_1_RIO')
28%|██▊       | 985.0/3496.9 [00:35<00:51, 48.31M/s]
>>> archive_path