src.eval.torf_rfvsnn_evaluation

Classes

TORFRFvsNNEvaluation(**kwargs)

class src.eval.torf_rfvsnn_evaluation.TORFRFvsNNEvaluation(**kwargs)
Author:

Alberto M. Esmoris Pena

Evaluation result comparing Random Forest (RF) and Neural Network (NN) predictions in the TransfOctoRF pipeline.

Holds the data necessary to generate a TORFRFvsNNReport (CSV files), a TorfRFvsNNPlot (multi-panel figure), and an output point cloud with RF, NN, and final predictions, class ambiguities, probabilities, and their differences.

Variables:

  • rf_proba (np.ndarray) – RF pseudoprobabilities (S, n_c).

  • nn_proba (np.ndarray) – NN probabilities (S, n_c).

  • final_proba (np.ndarray) – Final probabilities (S, n_c).

  • final_preds (np.ndarray) – Final predicted labels (S,).

  • y_true (np.ndarray) – Ground-truth labels (S,).

  • class_names (list or None) – Names for each class.

  • centroids (np.ndarray or None) – Centroid coordinates (S, 3).

__init__(**kwargs)

Initialize a TORFRFvsNNEvaluation.

Parameters:

kwargs – Evaluation attributes including data arrays.

report(**kwargs)

Transform the evaluation into a TORFRFvsNNReport.

See TORFRFvsNNReport.

Returns:

The report.

Return type:

TORFRFvsNNReport

can_report()

See Evaluation and evaluation.Evaluation.can_report().

plot(**kwargs)

Transform the evaluation into a TorfRFvsNNPlot.

See TorfRFvsNNPlot.

Keyword Arguments:

  • path (str) – The path to store the plot.

  • show (bool) – Boolean flag for showing the plot (True) or not (False).

Returns:

The plot.

Return type:

TorfRFvsNNPlot

can_plot()

See Evaluation and evaluation.Evaluation.can_plot().

write_pcloud(path, point_coords=None, point_labels=None, nearest_idx=None, include_prediction=True, include_class_ambiguity=True, include_probabilities=True)

Write a point cloud (LAS/LAZ) with RF, NN, and final predictions, class ambiguities, probabilities, and their differences (NN - RF).

When point_coords and nearest_idx are given, centroid-level values are propagated to the original points via closest-centroid assignment. Otherwise the point cloud is written at centroid level.

Parameters:

  • path (str) – Output file path.

  • point_coords (np.ndarray or None) – Original point coordinates (N, 3).

  • point_labels (np.ndarray or None) – Original point labels (N,).

  • nearest_idx (np.ndarray or None) – Centroid index for each point (N,).

  • include_prediction (bool) – Include predicted labels.

  • include_class_ambiguity (bool) – Include class ambiguity.

  • include_probabilities (bool) – Include class-wise probabilities.