IMPLEMENTATION OF ANIMAL FACE’S RECOGNITION BY CONVOLUTIONAL NEURAL NETWORK (CNN) ALGORITHM

The purpose of this research is to apply the Convolutional Neural Network (CNN) method in the field of Computer Vision. The CNN algorithm is a combination of Neural Network and Multilayer Perceptron which uses a convolution approach to extract features. The CNN technique was used to identify an animal dataset that has 16,130 images divided into three categories: Cats, Dogs and Wild. This study aims to recognize facial images of animals belonging to the category of Cats, Dogs or Wild Animals which resemble the derivatives of cats or dogs such as Lions, Tigers, Hyenas, Wolves and so on. Comparing to learning rate and epoch, the results are 10e-4 and 60 respectively. Utilizing random images from the datasets, learning rate and epoch may achieve an accuracy of about 97.22% or 116.33 out of 120 images. When using images taken outside of the datasets, the accuracy may be as high as 77.78% or 93.33 out of 120 images