Performance of Random Forest and SVM in Face Recognition
Emir Kremic, Abdulhamit Subasi
Faculty of Engineering and Information Technologies, International Burch University,
Bosnia and Herzegovina
Abstract: In this study, we present the performance of Random Forest (RF) and Support Vector Machine (SVM) in facial recognition. Random Forest Tree (RFT) based algorithm is popular in computer vision and in solving the facial recognition. SVM is a machine learning method and has been used for classification of face recognition. The kernel parameters were used for optimization. The testing has been comportment from the International Burch University (IBU) image databases. Each person consists of 20 single individual photos, with different facial expression and size 205×274 px. The SVM achieved accuracy of 93.20%, but when optimized with different classifiers and kernel accuracy among all was 95.89%, 96.92%, 97.94%. RF achieved accuracy of 97.17%. The approach was as follow: Reads image, skin color detection, RGB to gray, histogram, performance of SVM, RF and classification. All research and testing which were conducted are with aim to be integrated in mobile application for face detection, where application can perform with higher accuracy and performance.
Keywords: SVM, random forest, face recognition.
Received November 3, 2013; accepted October 26, 2014
