A COMPARATIVE STUDY OF DIFFERENT TYPES OF CNN AND HIGHWAY CNN TECHNIQUES

Abstract

In recent years, convolutional networks have shown breakthrough performance in image classification and detection. The main reason behind the performance of convnets is that they are inspired from the mammal’s visual cortex. In this paper, we have investigated the performance of four models that are Alexnet, Highway Convolutional Neural Network, Convolutional Neural Network and an evolutionary approach on highway convolutional neural network on the basis of train loss, test loss, train accuracy and test accuracy. These models are tested on two datasets that are WANG dataset and Simpsons dataset. In WANG dataset, Alexnet model achieved the highest test accuracy of 0.2625 and the highest train accuracy of 0.2193. Evolutionary Highway CNN has the least train loss of 0.1599 and CNN has the least test loss of 0.1604. In Simpsons dataset, Evolutionary Highway CNN has the Highest test accuracy of 0.5780. Highway CNN has the highest train accuracy of 0.5662 and for loss domain; Evolutionary Highway CNN has the least train and test loss