Application of deep learning for classification of two-dimensional radiographs

Abstract

Artificial intelligence (AI) has been a growing trend lately. One of the tasks which can be achieved by AI is computer vision, which is the ability for computers to process and analyse images, aiming to mimic human vision. One of the main tasks of computer vision is image classification, which is the process of labelling images into “classes”. For example, if there are images of multiple objects, and those images need to be categorized into “classes”, for instance “car”, “plane”, “ship”, or “house”, that is image classification. One common way to execute image classification is through convolutional neural networks, a technique implementing deep learning, which is a subset of machine learning, which is in turn a subset of AI. The objective of this thesis was to study the application of deep learning in image classification using convolutional neural networks. The Python programming language with the TensorFlow framework and Google Colaboratory hardware were used for the thesis. Models were chosen from available ones online and adjusted by the author. After the research, an accurate and performant model was developed and there is still room for further optimization.