Modelo de reconocimiento de rostros utilizando inteligencia artificial. Caso de estudio: Supervisión remota de exámenes en línea .

Abstract

In the context of the COVID-19 pandemic, the virtualization of traditional schools and hence online evaluation became a necessity to ensure the continuity of educational processes. However, identity fraud became a vulnerable aspect in remote environments, as it can compromise the validity and reliability of student knowledge assessment. In this thesis, a facial recognition model based on artificial intelligence algorithms is proposed to improve security in online evaluations. The main objective is to compare and select three facial recognition algorithms to choose the one that performs best in terms of accuracy and efficiency and meets the study's needs. To achieve this, three facial recognition algorithms were compared: Eigenfaces, Facerecognition, and FaceNet, the latter being a Convolutional Neural Networks model. Experiments were conducted using the faces of 10 students in various circumstances to evaluate the accuracy and efficiency of the algorithms. The results indicated that the FaceNet algorithm presented the best precision and efficiency in facial identification, surpassing the other two algorithms compared. However, the Face-recognition algorithm was chosen for implementation in the prototype due to its ease of use and being open-source software. The facial recognition prototype was developed using the Python programming language and the OpenCV library. Efficiency tests were conducted, and acceptable results were obtained for online application. The results indicate that the facial recognition prototype based on the Facerecognition algorithm is capable of identifying people accurately and efficiently in various circumstances. This demonstrates its usefulness in taking online evaluations to prevent identity fraud. This thesis contributes to the development of technological solutions for improving security in online evaluations. It also opens up the possibility of future research for implementation in other contexts, such as real-time security and surveillance systems for identifying people.