%PREX_SILHOUETTE_CLASSIFICATION PRTools introductory example
%
% Presents the construction of a dataset from a set of images,
% builds a classifier and performs an evaluation
%
help prex_silhouette_classification
delfigs
echo on
% load the Kimia image collection
a = kimia;
show(a,18);
%
% Select 5 classes and show them
b = seldat(a,[1 3 5 7,13]);
figure; show(b);
%
% compute the area
area = im_stat(b,'sum');
%
% compute the perimeter
bx = abs(filtim(b,'conv2',{[-1 1],'same'}));
by = abs(filtim(b,'conv2',{[-1 1]','same'}));
bor = or(bx,by);
figure; show(bor);
perimeter = im_stat(bor,'sum');
%
% construct a dataset with equal class priors and label the features
c = prdataset([area perimeter]);
c = setprior(c,0);
c = setfeatlab(c,char('area','perimeter'));
%
% Show the 2D scatterplot
figure; scatterd(c,'legend');
%
% Compute a linear classifier and show it in the scatterplot
w = ldc(c);
plotc(w,'col');
showfigs
%
% Compute and print classification errors
train_error = c*w*testc;
test_error = crossval(c,ldc,2);
echo off
fprintf('\nerror in training set: %5.3f\n',train_error)
fprintf('2-fold crossvalidation error: %5.3f\n',test_error)