TRndForest

The class TRndForest declares a random forest based classification engine. It actually implements multiple random forests, one for each class. The random forests are trained to distinguish one class against all other classes.

Properties

Methods

 

How To: Please follow these steps to train a classifier:
  1. create an instance of the class TRndForest
  2. set the basic parameters NTrees and RPar
  3. load a spectral descriptor set (SpdcSet.LoadUnchecked)
  4. load the training data using LoadTrnData
  5. call CalculateModel
  6. store the trained classifier using SaveOnDisk

 

How To: How to classify a dataset:
  1. load the data to be classified into a TMat4D structure (or, alternatively use the currently loaded dataset)
  2. load the trained classifier
  3. call ClassifyCubeData (Hint: for using the currently loaded dataset, set the parameter Cube to RawData and the parameter Metadata to MData).
  4. the classified data are available via ClassifiedCubeData