Skip to content

Related Packages

There are a number of other packages that combine Scikit-Learn with Xarray and/or use estimators for mapping. The table below attempts to compare their features (to the best of our understanding) with sklearn-raster so you can choose the right tool for your application.

Supported features are described in detail below.

Supported features sklearn-raster sklearn-xarray dask-ml scikit-map pyimpute pyspatialml
2D outputs ✅ ❌ ❌ ✅ ✅ ✅
3D outputs ✅ ❌ ❌ ✅ ❌ ❌
n-D outputs ✅ ❌ ❌ ❌ ❌ ❌
Preserves metadata ✅ ❌ ❌ ✅1 ✅1 ✅1
Parallel computation ✅ ✅2 ✅ ✅ ❌ ❌
Lazy computation ✅ ✅3 ✅ ❌ ❌ ❌
Xarray support ✅ ✅ ✅ ❌ ❌ ❌
Parallel fitting ❌ ✅ ✅ ❌ ❌ ❌
Custom estimators ❌ ✅ ✅ ❌ ❌ ❌
Raster processing tools ❌ ❌ ❌ ✅ ❌ ❌

Supported Features

  • 2D outputs: Package is capable of generating 2D model outputs from 2D feature inputs, e.g. predicting georeferenced land cover maps or transforming geospatial imagery.
  • 3D outputs: Package is capable of generating 3D model outputs from 3D feature inputs, e.g. predicting a time series of georeferenced maps.
  • n-D outputs: Package is capable of generating model outputs with arbitrary dimensionality from n-D feature inputs, e.g. predicting a time series of climate variables at various pressure levels.
  • Preserves metadata: Model outputs retain the metadata of the input data, such as spatial references, band names, and NoData masks.
  • Parallel computation: Package computes model outputs in parallel.
  • Lazy computation: Package computes model outputs lazily, deferring computation until necessary.
  • Xarray support: Package supports Xarray data structures as model inputs and outputs.
  • Parallel fitting: Package supports fitting estimators in parallel to reduce training time.
  • Custom estimators: Package implements additional estimators.
  • Raster processing tools: Package includes additional functionality for processing raster data beyond ML modeling.

  1. scikit-map, pyimpute, and pyspatialml only preserve metadata when writing outputs directly to disk. ↩↩↩

  2. sklearn-xarray supports parallel operations by wrapping dask-ml estimators. ↩

  3. sklearn-xarray supports lazy evaluation by deferring computation, but requires that the entire dataset is loaded into memory. ↩