Not that instead of using output of one application as input. I used Baseline data of the first application as input of the second one. It does not strickly break the dependency between the tests but avoid to push new images to the OTB-Data repository.

ENH: Update the constructors of all the available classifiers with the OpenCV default parameters in TrainImagesClassifier

BUG: Inversion of parameters in the constructor of the GBT classifier in the TrainImagesClassifier application

TEST: Add tests of several classifiers for the OpenCV framework in the TrainImagesClassifier application

ENH: Replacing the 3 SVM classification applications to handle the OpenCV machine learning framework

ENH: Removal of the numbers of reference and produced labels input parameters in the otbComputeConfusionMatrix application

ENH: Definition of a format for confusion matrix files in otbComputeConfusionMatrix and otbFusionOfClassifications applications

TEST: apTvFeLineSegmentDetectionNoRescale test new parameter: -norescale true, the baseline was consequently updated

ENH: ClassificationMapRegularization Application: structuring element (ball) with the same radius along X and Y, the corresponding documentation, test and baselines were updated.

ENH: Classification Map Regularization by Majority Voting, new parameter names in accordance with itkLabelVotingImageFilter

ENH: Majority Voting: multiple majority labels are set either to UnDefinedValue or to their original value (by default)