Pet Face Detection

Abstract

The technology of human face recognition is very useful and important for man-machine communication and security. Then, many methods of face recognition have been developed and are used for a lot of systems such as digital camera, bank ATM (Automated Teller Machine), home security and so on. These days, most of human faces are precisely recognized with Haar-like features and AdaBoost classifiers. On the other hand, many people have pets such as dogs and cats, and they are friends or even family members for human beings. Then, surveillance cameras have been developed and used for monitoring pets, who are in houses while the owners are away from home. However, the recognition of pet faces is very difficult since there are a variety of patterns on the face. It is not sure that pet faces can be recognized with the same methods as the faces of human beings. Therefore, we have investigated the methods to detect pet faces, especially for cats and dogs. We have used Haar-like and HOG (Histogram of Oriented Gradients) features to construct the classifiers of pet faces. In addition, we have tried to investigate which is better for the pet face detection to use either whole face or parts of the face such as eyes and nose. Finally, we have applied our method to some pictures of cats and dogs, and confirmed that the classifiers have been able to detect the pet faces correctly.