Image Classification of Canaries Using Artificial Neural Network
Bagus Yanuki, Aviv Yuniar Rahman, Istiadi Istiadi
Abstract
The canary is a famous animal in Indonesia. Therefore, many in Indonesia are maintaining and cultivating canaries. However, because of the various types of canaries, the general public often misinterprets canaries. Therefore, the researcher proposes canary image classification system using Artificial Neural Network based on texture, shape and color features. The test process in comparison uses 3 methods, namely Naive Bayes, SVM with 4 variations of NU-SVC Linear, NU-SVC Polynomial, NU-SVC Radial, NU-SVC Sigmoid and Artificial Neural Network. Results starting from Naïve Bayes maximum value of 65% split ratio of 90:10. The SVM NU-SVC Linear variation maximum value of 60% split ratio of 90:10. Variation NU-SVC Polynomial maximum value of 43% split ratio 90:10. The NU-SVC Radial variation maximum value of 60% split ratio 90:10. The NU-SVC Sigmoid variation maximum value 42% split ratio 90:10. The Artificial Neural Network method maximum accuracy value 96% split ratio 90:10 between training data and testing data. This test shows that the Artificial Neural Network can classify the image canary species based on the level of texture, shape and color. So that it can be easier to distinguish by finding the accuracy value based on the learning rate of various types of canary colors.