Bagging de Detectores por Produto Interno para Detecção de Olhos

This article presents a system for eye detection using an ensemble of correlation-based filters known as Inner Product Detector (IPD). This work has three main contributions: i) an ensemble classifier with higher accuracy than the original IPD detector, using the bagging algorithm; ii) new discriminant functions based on the ensemble output; and iii) the study of the influence of bagging on the system performance. The proposed method was evaluated on the BioID dataset, achieving an average accuracy of 98.02% and 95.71% for right and left eyes, respectively, where a deviation of up to 10% of the interocular distance is considered a hit.