Eureka! Institutional Repository

In this master thesis, a review will be made of the evolution of machine learning and neural networks. Convolutional neural networks (cnn), belong to the broader spectrum of deep learning neural networks. Their main features will be presented and the way they are implemented will be analyzed, as well as their sections and architectures. Synchronous neural networks are very often applied to image and video analysis problems, which means that their use is necessary for the development of an application for automatic emotion recognition through images or videos of persons. Such applications have critical success factors and challenges, which will also be presented. The aim of this work is to develop and evaluate deep learning models that will be used in problems of automatic expression recognition. This relatively new research field of image analysis, consists of such problems and focuses on models of machine learning that learn to interpret facial expressions. The training of these models will be done on the data collection ‘facial emotion recognition 2013’ which consists of greyscale images of faces. Finally, an application will be developed that will use the models created to recognize real-time expressions using the camera of any conventional computer.