Add Great African Food Company Crop Type Tanzania Dataset

[nilsleh]

This PR adds the Crop Type Tanzania dataset from Radiant MLHub.

It features time series data with polygon crop type annotations as segmentation masks. I have implemented it as a GeoDataset to allow for train/val/test splits based on geographical location and am looking to do that with other datasets of this type (also converting CV4A_Kenya_Crop_Type dataset to a GeoDataset). Following the TODO in cv4a_kenya_crop_type.py this implementation is populating the rtree index by using stac.json files. The __getitem__ method returns a tensor with dimensions time x num_bands x height x width. This dataset has both rasters as input, as well as vector annotations, so sort of a hybrid between RasterDataset and VectorDataset.

Dataset Features:

• 392 annotations with 6 different crop type classes for 44 different labeled areas that each have a variant amount of time series inputs in form of Sentinel 2 imagery

Dataset Format:

• separate sentinel 2 bands as tif file as well as a cloud probability layer (images in epsg 32736)
• stac.json files for each input image tile (bboxes in epsg 4326)
• geojson files with polygon annotation and label (polygon coordinates in epsg 32736)
• stac.json for labels

Issues:

• I am not creating the correct dummy data, something is off with the bounds, and therefore tests are
• here aren't many annotations and the ones that are there are all very small (see below for example) making me question whether I mess up the indexing when creating a segmentation mask from the polygon annotations
• there is also another design choice regarding the included datetime in stac.json files: if this datetime is used to populate the index, then RandomGeoSampler will also sample time instances and a given bounding box query will not return all timesteps for a given geographical XY location, as it is maybe expected.

Example

:

[adamjstewart]

and am looking to do that with other datasets of this type (also converting CV4A_Kenya_Crop_Type dataset to a GeoDataset)

❤️

[adamjstewart]

Superseded by #512