This operator offers the possibilty to train or validate a support vector machine model for classifying scratch assay images into such containing a scratch and such that don't (already closed)
newly created models can be used in the Scratch Assay Analyzer
In order to use this operator two directories have to be created, one containing images where there is a scratch visible (positive samples) and a second containing images without a scratch (negative samples)
all images should have the same dimensions and the scratches have to be oriented either all horizontally or vertically
directory containing positive samples
directory containing scratch assay images where the scratch has not been closed
directory containing negative samples
directory containing scratch assay images where the scratch has already been closed
scratch orientation
horizontally or
vertically
entropy filter size
size of entropy filter mask
increase lets the scratch area decrease
sigma
standard deviation of gauss filter
increase leads to more image smoothing and scratch area tends to decrease
regularization parameter (C)
determines severity of misclassification of outliers
kernel type
determines type of kernel for support vector machine
available options are LINEAR, RADIAL and POLYNOMIAL
maximum iterations
maximum number of iterations for level set segmentation
degree
degree of polynomial if polynomial kernel is chosen
k
fold number if k-fold cross validation is performed
validation method
which (if any) validation method shoulb be used
available options are
NONE: train a new svm model
K FOLD: stratified k-fold cross validation
LEAVE ONE OUT: leave-one-out cross validation
verbose
output somme additional information