Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • master default protected
1 result
Created with Raphaël 2.2.027Jan262522201915141312118723Dec2117161514119427Nov26259Oct52128Sep2523211817161514111098432131Aug28272625242120191817141312111030Jul29282724232221201716151410987632Improved extract by mask utility.Improved the function to replace array values such that it handles also string datatypes.Implemented the computation of the spectral distribution of the different classes.Added seaborn dependency.Improved extraction from raster: implemented clip to shapefile.Updated dependencies.Adjusted NetworkInference to accumulate statistics accross different model runs.Adjusted imports to match new naming.Split the configuration files for different types of training and evaluation.Working on a more stable generation of model state files.Moved random seed parameter to split configurations.Implemented cross-validation subsampling.Training script to train models on a single domain.Preparing to split transfer learning to dedicated training script.Major refactor of NetworkInference: implemented label mapping.Adjusted dataset source directory.Compliance to gdal v3.2 python bindings.Removed some options for network evaluation.Major refactor: Adapting NetworkInference to accomodate multiple models.Removed the date option for plot_sample.Added LabelMapping from Sparcs to Alcd and vice versa.Handle 2d-input arrays.Moved transfer learning part to dedicated configuration class.Close gdal vrt dataset for safety reasons.Changed naming convention.Improved representation logging.Fixed some issues when initializing model training.Improved documentation of available options.Moved training initialization to dedicated script.Fixed band order on both Windows and Linux.Improved logging when converting hdfs to tifs.Implemented to set NoData value when saving numpy arrays to tifs; implemented band description in vrt files.Major refactoring towards a general training class for classification problems.Implemented changes of the configurable parameters.Added UDA support at input feature level.Number of input channels and output classes are saved to state file; configurable loss function support dropped.Decluttered checkpoint/transfer learning model initialization.Major generalization of Network class; added support for AdamW and AMSgrad.Implemented configuration of optimizer keyword arguments.Added a function to calculate the extent of a low-resolution pixel in a high resolution image.