Dog Breed Classifier for Facial Recognition using Convolutional Neural Networks
Bickey Kumar Shah, Aman Kumar, Kumar Amrit
Abstract
This paper dealt with the breed classification of dogs. To classify dog breed is a challenging part under a deep convolutional neural network. A set of sample images of a breed of dogs and humans are used to classify and learn the features of the breed. The images are converted to a single label of dimension with image processing. The images of human beings and dogs are considered for breed classification to find the existing percentage of features in humans of dogs and dogs of human. This research work has used principal component analysis to shorten the most similar features into one group to make an easy study of the features into the deep neural networks. And, the facial features are stored in a vector form. This prepared vector will be compared with each feature of the dog into the database and will give the most efficient result. In the proposed experiment, 13233 human images and 8351 dog images are taken into consideration. The images under test are classified as a breed with the minimum weight between test and train images. This paper is based on research work that classifies different dogs breed using CNN. If the image of a dog is supplied then the algorithm will work for finding the breed of dog and features similarity in the breed and if the human image is supplied it determines the facial features existing in a dog of human and vice-versa.