Concept Detection using Multiple Feature Set and Classifiers

Abstract

Visual concept detection is the task of determining concept present in image or video by extracting low level features and training of classifiers in general. Researchers have used various features and classifiers for concept detection. In this paper performance evaluation of fusion of features and classifier is presented. Color moment, HSV histogram, wavelet transform and combination of these features have been used in proposed system. Artificial Neural Network (ANN) and Support Vector Machine (SVM) are employed for classification. The proposed system is implemented on Corel 1K image dataset and Trecvid 2007 benchmark video dataset. The system performance is evaluated using predictive measures of precision, recall and f score. Using simple fusion of features average precision of SVM classifier is better than ANN. The proposed global feature fusion based method is simple yet effective in concept detection task.