ESTABLISHMENT OF DEEP LEARNING PERFORMANCES FOR CREDENTIAL OF SPECIFIC BIRD PATTERNS USING CURRENT TRENDS

Abstract

Birds have accompanied human race from the evolution throughout the period of existence in many ways. Birds can be classified according to their purpose as food, for recreation, as pet and an inspiration in ancient and modern literature. Birds form a predominant class in the animal kingdom sharing certain features like feathers, wings, flight speed and migration which differentiate among them. Nearly there are 10,000 varieties in different sizes, characters and enormous ranges spreading all over the world including equable regions like Antarctica. So classification of birds became inevitable criteria for researchers and scientists to study about bird types. Ornithologists rely on numerous techniques based on machine learning and deep learning approaches to study about bird categorization. This paper focuses on the classification of birds by deep learning methods. For this purpose most famous variety of birds are collected from the data set and the initial step namely pre- processing is done with the sample of birds. Classification is done by three popular techniques namely CNN model, InceptionV3 and Mobilenet V2 method. The highest accuracy nearly 99% is obtained from Mobilenet model and the performance is compared with other two methods.