This release includes the following new and improved features.

Highlights

This release contains the following changes:

  • New model extraction tasks for Deep Learning and Machine Learning.

  • A new property has been added to all Machine Learning supervised classification tasks.

New and Updated Tasks

These new tasks have been added to Deep Learning:

  • ExtractDeepLearningGridModelFromFile: Opens a deep learning grid model from a file.
  • ExtractDeepLearningObjectModelFromFile: Opens a deep learning object model from a file.

This new task has been added to Machine Learning. Machine Learning documentation can be found under ENVI Machine Learning in the ENVI Help contents.

  • ExtractMachineLearningModelFromFile: Opens a machine learning model from a file.

  • In Machine Learning 3.0, normalization was handled differently than in it was in version 2.0. The default was to use histogram stretching to normalize the data per the data’s min and max values. A new REMOVE_OUTLIERS property has been added to the following tasks. This property disables histogram stretching so that normalization is performed how it was done in Machine Learning 2.0.

    • NormalizationStatistics
    • TrainExtraTrees
    • TrainKNeighbors
    • TrainLinearSVM
    • TrainNaiveBayes
    • TrainRandomForest
    • TrainRBFSVM

Updated Tools

Deep Learning tool updates:

The Patches per Batch parameter has been added to the Train Deep Learning Pixel Segmentation Model dialog.

Machine Learning tool updates:

The Remove Outlier parameter has been added to the following Machine Learning tools. See the description under New and Updated tasks for details.

  • Extra Trees Classification
  • K Neighbors Classification
  • Linear SVM Classification
  • Naive Bayes Classification
  • Random Forest Classification
  • RBF SVM Classification
  • Machine Learning documentation can be found under ENVI Machine Learning in the ENVI Help contents.