Remote Sensing Processor 0.2.2

This update reworks semantic segmentation functions and improves processing speed and stability

• Now uses xarray, dask and rioxarray instead of numpy and rasterio
• Now stores tiles in zarr containers instead of hdf5
• Syntax, inputs and outputs of all segmentation functions are reworked
• Custom kwargs can be used when initialising models
• Augmentations can be applied while training with augment arg
• Dataset size can be increased by repeating it n times while training with repeat arg
• Raster histograms now can be matched while creating mosaic using match_hist arg
• Specific value in a raster can be replaced using replace_value function
• Nodata value in a raster can be replaced using replace_nodata function
• Vector file can be rasterized using rasterize function
• Sentinel2 now can be upscaled using resampling algorithm. superres arg is renamed to upscale, resample arg added
• Sentinel2 now can be normalized using normalize arg
• Landsat thermal bands now can be normalized using normalize_t arg
• clipper argument is renamed to clip
• projection argument is renamed to crs

Remote Sensing Processor 0.2.1

• Added normalize function that applies min/max normalization to data
• Segmentation train and test now support multiple datasets input
• Segmentation train, test and generate_map now support multiprocessing
• Added support for more Landsat products
• Various bug fixes

Remote Sensing Processor 0.2

Remote Sensing Processor 0.2 adds image segmentation module

• Added train and test functions that train and test pytorch and sklearn segmentation models
• generate_tiles and generate map functions reworked and moved to rsp.segmentation module
• Sentinel-2 superresolution algorithm rewritten in pytorch
• normalized_difference function renamed to calculate_index

Remote Sensing Processor 0.1

Remote Sensing Processor 0.1

This is the first release of Remote Sensing Processor.

It includes Sentinel-2 and Landsat preprocessing, creating raster mosaics, calculating normalized difference indices (for now NDVI only), cutting rasters to tiles and creating maps using pre-trained models.