Role of Filter Sizes in Effective Image Classification Using Convolutional Neural Network

Abstract

Over the past few years, Deep Neural Networks have provided us the best results on a variety of problems, such as pattern recognition, computer vision, and Speech recognition and image classification. Convolutional neural networks are one of the deep learning models which are mostly used in image classification and are the base of many other deep neural network models. Convolution neural network uses convolution and pooling layers for feature abstraction. Unlike a regular Neural Network, the layers of a Convolutional neural Network have neurons arranged in three dimensions: width, height, depth, and filter sizes of different dimensions are used for feature reduction. But the problem with the convolution neural network is that it is difficult to train and can led to overfitting. There are many factors to look for while designing Convolutional Neural Networks one of them is filter size. Dimensions of filter sizes play a very important role for effective training in the convolutional neural network, So, here in this paper we compared the results of 3 × 3, 5 × 5 and 7 × 7 filter sizes and checked training accuracy, test accuracy, training loss, test loss as constraints.