Detección de emociones y reconocimiento facial utilizando aprendizaje profundo

Abstract

With technological advances in the area of artificial intelligence, and especially thanks to deep learning techniques, image classification and facial recognition are possible. This project has created a deep learning architecture with four layers for the detection of emotions, using five classes angry, happy, neutral, sad and surprised of the Kaggle dataset called Learn facial expressions from an image FER2013 , using a total of 15,000 images for the training of the convolutional neural network in the Kaggle Notebooks platform, the tools used were the Python language, the Keras library, Tensorflow, OpenCV and the pre-trained algorithm Front Face of Haar Cascade, obtaining an accuracy of 67.13. While the Face Recognition library was considered for facial recognition, it has an accuracy of 99.38. Finally, a web service based on Django was developed to respond to requests made from a web application, implemented in Laravel, for the communication of the two frameworks the axios library was used, access to the application is made through face recognition, once the session is started, it is possible to perform tests for the detection of emotions and facial recognition.