Version 0.4.1

Change log:

• Added eolearn.features.radiometric_normalization module with EOTasks: ReferenceScenes, HistogramMatching, BlueCompositing, HOTCompositing, MaxNDVICompositing, MaxNDWICompositing and MaxRatioCompositing. Contributed by @JohannesSchmid
• Jupyter notebooks about LULC restructured and updated.
• Implemented eolearn.core.FeatureTypeSet class.
• eolearn.core.FeatureParser has new parameter allowed_feature_types.
• Fixed masking in InterpolationTask.
• Added MaskFeature task.
• Binary masks in eolearn.io and eolearn.mask are now cast to numpy.bool dtype.

Version 0.4.0

Change log:

• Core changes:

  • Support for Python 3.7 added.
  • EOPatch now supports vector features in geopandas.GeoDataFrame format and makes checks that they can't be added in any other format.
  • EOPatch checks that each discrete feature type has only numpy arrays with integer or bool dtype. If that is not the case a DeprecationWarning is thrown.
  • EOPatch now checks correct form of bbox and timestamp feature types. BBox has to be of type sentinelhub.BBox and timestamp a list of datetime.datetime or datetime.date objects.
  • Better error propagation while executing EOWorkflow or EOExecutor
  • Packages attrs, geopandas and sentinelhub are now core dependencies of eo-learn.
  • EOExecutor parameter file_path replaced by logs_folder
  • Improvements of EOExecutor multiprocessing functionalities
  • Parameters of EOWorkflow.dependency_graph method changed.
  • Minor improvements and fixes.

• Other changes:

  • Added KrigingInterpolation task to eolearn.features - contributed by @jO-Osko
  • Added ErosionTask and RasterToVector task to eolearn.geometry.
  • Minor improvements and fixes of various tasks.

Version 0.3.3

Change log:

• Added example notebook with a complete processing chain of land use/land cover classification
• Improvements of eolearn.features.InterpolationTask
• Added eolearn.io.S1IWWMSInput and eolearn.io.S1IWWCSInput tasks - contributed by @wouellette

Version 0.3.2

Change log:

• Improved saving functionalities of EOPatch:
  • overwrite parameter replaced by overwrite_permission which can be either OverwritePermission.ADD_ONLY or OverwritePermission.OVERWRITE_FEATURES or OverwritePermission.OVERWRITE_PATCH.
  • compress parameter removed, uncompressed data can be saved by compress_level=0.
• Added BlobTask, HOGTask and LocalBinaryPatternTask tasks to eolearn.features - contributed by @HugoFRN
• Improvements of eolearn.features.InterpolationTask. Parameter resample_range can now also receive a list of new dates.
• Fixed issue with EOExecutor missing a report.html file.
• Various minor improvements and docs updates.

Version 0.3.0

First official release of the package. Recent changes:

• major improvements of all basic elements in eo-learn-core subpackage (i.e. EOPatch, EOTask, EOWorkflow and EOExecutor),
• various new EOTasks added,
• multiple EOTasks were renamed or their initialization parameters were changed,
• documentation released (link),
• package released to PyPI and can now be installed with pip (instructions).