QLearn
QLearn preforms automatic training of a neural network model on raster data
QLearn is a plugin that allows you to train a UNet segmentation neural network architecture to segment and classify raster data. It can also use pretrained models to make predictions on raster data
Required Dependencies: torch and torchvision
Install with OSGeo4W Shell (pip3 install torch torchvision)
QGIS Integration: Train and use machine learning models without leaving QGIS
Automatic Data Preprocessing: Automatic alignment, rescaling, reclassifying, and normalization of rasters
Testing and Validation: Automatically splits datasets into training and evaluation sets and provides model evaluation metrics
Confidence Filtering: Optional filtering of predictions based on confidence levels
Multiband Support: Works with any number of input bands and output classes