Sentinel2 Cloud Detector Versions Save

Changelog:

• restrict numpy version to < 2 to avoid issues with the upcoming numpy 2.0 release

Changelog:

• Removed Python 3.7 support (due to sentinelhub-py changes) and add official Python 3.11 support.
• Evalscript generation is now done by utilities from sentinelhub-py, the functions get_s2_evalscript and get_timestamps were removed.
• Removed scipy and scikit-image from dependencies.
• Switched to a pyproject.toml based installation and exclude unnecessary files from builds.
• Various other improvements.

Changelog:

• (codebreaking) Removed CloudMaskRequest class. It's value was extracted to utils.get_timestamps and utils.get_download_bands_and_valid_data_mask. See example on how to adjust.
• (codebreaking) Methods of S2PixelCloudDetector now enforce a stack of images as input and no longer auto-adapt for single images.
• The package is now fully annotated and type-checked.
• Improvements to dtypes of output to reduce memory footprint.
• Overhaul of test suite.
• Various minor improvements.

Changelog:

• Improved how S2PixelCloudDetector.get_mask_from_prob handles single-temporal images

Changelog:

• Switched to Catalog 1.0.0. introduced with sentinelhub-py 3.8.0

Changelog:

• Removed support for Python 3.6, added official support for Python 3.9 and 3.10.
• PyPI package release now also includes a wheel.
• Fixed a deprecation warning raised by newer versions of sentinelhub package.
• Applied black and isort formatters.

Changelog:

• Refactored CloudMaskRequest class and its parameters. Now it is collecting data from Sentinel Hub Process API instead of OGC. It downloads data in digital numbers and it is optimized for performance and download costs.
• Updated the tutorial notebook on how to use s2cloudless.
• Removed support for Python 3.5.
• Restructured the internal structure of the content. The names of modules have changed.
• Updated and improved unit and integration tests.

Change log:

• The cloud detection model now loads lazily when first needed instead of right away at class instance initialization.
• The cloud detection model is now being loaded natively with lightgbm.Booster instead of sklearn.external.joblib.
• Removed scikit-learn dependency.

Change log:

• Removed test_cloud_detector.py module from the package. This reduces package size and dependencies. The same tests can still be run from tests folder in GitHub repo.

Change log:

• CloudMaskRequest now automatically sets correct image format, disables atmospheric correction and overrides layer definition.
• CloudMaskRequest does not send a duplicated HTTP request to WFS services anymore.
• Improvements in error handling.
• Better unit tests added.