Klasifikasi Hewan Mamalia Berdasarkan Bentuk Wajah Menggunakan Fitur Histogram of Oriented dan Metode Support Vector Machine

Abstract

Mammals have several characteristics that can be distinguished, such as footprints, voice, and face shape. Mammals can be recognized. To classify the face shape of mammals, the Histogram of Oriented Gradients (HOG) and Support Vector Machine (SVM) methods can be used. This study uses the LHI-Animal-Faces dataset which is taken as many as 15 species of mammals, where each type of mammal is selected 60 images and resized to 150x150 pixels. The image is converted into a grayscale image for the HOG feature extraction process. Furthermore, the classification process uses SVM. The kernels used are Linear, Polynomial, and Gaussian kernels. The testing process uses K-Fold Cross Validation. The folds used are 3-fold, 4-fold, 5-fold, 6-fold, and 10-fold. The performance of the HOG feature and the SVM method that gives the best results is the Linear kernel using 10-fold with an accuracy value of 96.55%, precision of 77.92%, and recall of 74.11%. The sequence of kernels that give the best results in this test is the Linear kernel, Polynomial kernel, and Gaussian kernel.

Keywords

• hog
• k-fold cross validation
• mammals
• svm