Development of a Face and Emotion Recognition System using Computer Vision and Transformers

Óscar Parro Sainz. (2025). Development of a Face and Emotion Recognition System using Computer Vision and Transformers. Final Career Project (TFM). Universidad Politécnica de Madrid.

Abstract:
Emotion detection refers to the automated recognition and identification of human emotions by enabling machine learning algorithms to analyze complex human patterns. Recent advancements in algorithms capable of extracting facial landmarks, biosignals, and other physiological or behavioral features have significantly advanced this field. These developments have expanded the applicability of emotion detection technologies across domains such as healthcare, marketing, and customer service. This project is particularly pertinent in the context of the increasing influence of the Internet, where high-quality information coexists with manipulative content and disinformation campaigns. By detecting emotions and analyzing facial responses, the system can help identify emotional manipulation in information sources and social networks, fostering greater critical awareness. In this way, it is aligned with initiatives that seek to promote responsible access to information, the development of democratic values and quality education. The main objective of the project is to develop an innovative system that, utilizing advanced computer vision techniques and deep learning models. Specifically, transformer-based architectures for real-time face detection and precise emotion recognition. The system is designed to monitor and analyze users’ emotional reactions to a variety of stimuli. To do this, it uses an interface that allows, firstly, to identify the person through facial recognition and, secondly, to detect their mood. Emotion detection and analysis using advanced computer vision techniques are promising solutions to this challenge. Automatic emotion recognition allows obtaining accurate information about users’ emotional reactions to specific content. This could help identify emotional patterns associated with manipulative or potentially misleading information, thus helping to promote more critical and responsible consumption of online information.