Summary of the classification results in cross-validation on the learning sets and on the independent test sets using the “All binary” or “Two-tier” method.

For each phenotype (column 1), the optimal mode of classification (“BAGS” or “C”) (second column), the size of the extracted random subwindows (expressed in percentage of the size of the original image) (third column) and the kind of test type used for each node (SIMPLETHRES or DIFFNEIGHBOR)(fourth column) are given. “CV rate on LS” gives the results for each phenotype obtained in cross-validation on the corresponding learning sets (LS) with these parameters in the following form: global recognition rate, recognition rate of the specific phenotype, recognition rate of the corresponding “negative” phenotype. “Rate on TS” gives the results obtained with the same parameters on the corresponding independent test sets, respectively with the “All binary” or with the two-tier approach (three-class model followed by binary classification). Recognition rates are in % of the corresponding set.
Summary of the classification results in cross-validation on the learning sets and on the independent test sets using the “All binary” or “Two-tier” method.