Emotions! Neural network for emotion recognition and backend applications

This master thesis focuses on neural networks and their practical deployment in a startup project EMOTIONS!. This project is a group work of five members, which was made possible by the Data Project Challenge 2023. The project is divided into several parts, which are linked to create a functional platform, both from a business and technical point of view. This thesis is divided into two parts. The first part defines the background definitions and introduces the world of neural networks by applying a qualitative analysis. Furthermore, this section focuses on the shortlisting of neural networks and their advantages in order to find the most suitable type of network to fulfill the business requirement. The second part builds on the results of the previous passage and uses them to achieve a robust model and its implementation in the platform. Within this section, the analysis of the datasets, the different iterations of the model architectures and their learning process are performed. Finally, the thesis concludes with the implementation of the most suitable model in the Google Cloud platform as a back-end, connecting it to the front-end and testing its functionality. The thesis provides an end-to-end view of creating a neural network and its practical deployment, and more importantly its potential use by a wide audience. As a result of the thesis, a key part of the EMOTIONS! platform providing emotion detection has been created, which could be realized thanks to parts of colleagues' work.