Convolutional neural networks in image recognition

Abstract

In this thesis, we study the topic of deep learning with a focus on image recognition using convolutional neural networks. We cover the various components of deep learning, including the network structure, backpropagation and stochastic gradient descent. We explain the fundamentals of these components and compare theory to practice. We then examine convolutional neural networks and the various layers they consist of. Finally, we build and train a convolutional neural network to classify small images of coloured shapes. This network achieved an accuracy of around 85%.