An Approach to Animal Face Identification and Classification System using CNN

Abstract

The Proposed work gives a significant contribution to classification and identification of animal faces in the low-resolution images using most commonly and widely used neural network system. We have adopted Convolutional Neural Network (CNN), whichgives promising results in classifying the images using neural network system. CNN helps to extract features of animal faces automatically, also helps to learn and classify them. For the proposed work, we have used the CIFAR-10 dataset (Canadian Institute For Advanced Research), more than 6000 images are used to identify and classify the animal faces. Comparing to other classification techniques,CNN is progressivemethodto extract features and classify dynamically, where as other conventional algorithms were hand-engineered which means programmer has to extract the features from the dataset explicitly. In our work, we have gained 92% accuracy in identifying and classification of the animal faces for low-resolution images.

Keywords:Animal classification, Threshold, Segmentation, Convolution Neural Network, Digital image processing, natural scene.