nets apply-model

Applies a trained keras model to a series of feature datasets to map model predictions.

---

usage: nets apply-model [-h] -f filepath -o output_file
                        [--fscaler features_scaler]
                        [--rscaler response_scaler] [--eval-window percent]
                        [--nodata-pass] [--nodata-fraction percent]
                        [--te xmin ymin xmax ymax] [--ot dtype]
                        [--dstnodata value] [--use-latest] [--cores CORES]
                        model_config

positional arguments:
  model_config          Path to the myco model configuration file (.yml). From
                        the saved model directory.

optional arguments:
  -h, --help            show this help message and exit
  -f filepath, --features filepath
                        Path to raster-format feature data to extract
                        (default: None)
  -o output_file        Path to the output output model predictions. (default:
                        None)
  --fscaler features_scaler
                        Path to the scaler to apply to the feature data.
                        (default: None)
  --rscaler response_scaler
                        Path to the scaler to apply to the response data.
                        (default: None)
  --eval-window percent
                        The proportion of the interior window to save
                        predictions from. (default: 0.5)
  --nodata-pass         Read the feature data prior to application to filter
                        out tiles with nodata. (default: False)
  --nodata-fraction percent
                        The proportion of allowable nodata in a tile to apply
                        the model. (default: 0.0)
  --te xmin ymin xmax ymax
                        Apply the model only to these target extent bounds.
                        Units must match features CRS. (default: None)
  --ot dtype            The output raster datatype. (default: float32)
  --dstnodata value     The output raster nodata value. (default: None)
  --use-latest          Apply the most recent model instead of the best model.
                        (default: False)