Improving Image Recognition Capacity in Convolution Neural Networks

Abstract

Creating accurate Machine Learning models capable of identifying and localizing multiple objects in a single image remained a core challenge in computer vision. But, with recent advancements in Deep Learning, Object Detection applications are easier to develop than ever before. Deep neural networks and deep learning have become popular in past few years, thanks to the breakthroughs in research, starting from AlexNet, VGG, Google Net, and ResNet. Recently, ResNet, was reported to greatly improve the performance of large-scale image recognition and helped increase the popularity of deep neural networks. This paper reports the results of an empirical study of methods and techniques for improving the capacity of Convolution Neural Networks to solve an image recognition problem. From the results, it was noted that optimization techniques greatly improve the performance of Convolution Neural Networks for image processing tasks. The performance measures employed in the study included Mean Squared Error Accuracy as well as convergence rate