add_ground_csf          Add ground as class 3 using CSF after
                        FuelDeep3D prediction
config                  Default config Create model configuration
ensure_py_env           Ensure a Conda environment and Python deps for
                        FuelDeep3D
evaluate_single_las     Evaluate a single LAS using custom field names
evaluate_two_las        Evaluate two LAS files (truth vs prediction)
install_py_deps         Install Python dependencies into a Conda
                        environment
palette_veg             Default vegetation color palette
plot_confusion_matrix   Plot confusion matrix as a heatmap (ggplot2)
plot_las_3d             3D Visualization of LAS Using RGL
predict                 Run prediction on cfg$las_path and write output
                        LAS If cfg$num_classes == 4, add ground using
                        CSF automatically.
print_confusion_matrix
                        Pretty-print confusion matrix
print_metrics_table     Pretty-print evaluation metrics table with
                        summary row
py_setup                Create/use a venv and install Python deps
                        listed in extdata/python/requirements.txt
remove_noise_sor        Remove sparse outlier points using Statistical
                        Outlier Removal (SOR)
train                   Train the FuelDeep3D model (auto-preprocess if
                        NPZ tiles are missing)
