An evaluation of divide-and-combine strategies for image categorization by multi-class support vector machines

Categorization of real world images without human intervention is a challenging ongoing research. The nature of this problem requires usage of multiclass classification techniques. In divide-and-combine approach, a multiclass problem is divided into a set of binary classification problems and then the binary classifications are combined to obtain multi-class classification. Our objective in this work is to compare several divide-and-combine multiclass SVM classification strategies for real world image classification. Our results show that One-against-all and One-against-one MaxWins are the most efficient methods.