TY - GEN
T1 - Performance Evaluation of Face Detection Algorithms for an Emotion Recognition Application in a School in the Department of Nariño - Colombia
AU - Díaz-Toro, Andrés
AU - Cervelión-Bastidas, Álvaro
AU - Campaña-Bastidas, Sixto
AU - Méndez-Porras, Abel
AU - Alfaro-Velasco, Jorge
AU - Jiménez-Delgado, Efrén
AU - Calvo-Valverde, Luis
N1 - Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2023
Y1 - 2023
N2 - Emotion recognition in digital images, based on the facial expressions of people, can add value in different areas such as education, shopping centers, hotels, entertainment centers, restaurants, among others, since it allows a better understanding of the requirements of the people, improve services, and predict sales trends. In a classroom, this technology allows to identify in real time the reaction of students to the development of the class, and in this way, the teacher can make the necessary adjustments to improve the learning process. The first step for this application is to detect faces of multiple students present in the scene, with efficient algorithms that process good-quality images. In this paper, the performance of six face-detection algorithms is determined using images taken in a classroom, in the town of Túquerres, in the department of Nariño, Colombia. The results show that a good camera resolution of 5 megapixels or higher, and good lighting conditions are determinant for successful face detection in classrooms of approximately 46 m2. In addition, the best performance was obtained with RetinaFace algorithm, which is more robust to different facial postures, achieving an accuracy of 96.5% with poor lighting conditions and 97.84% with good lighting conditions.
AB - Emotion recognition in digital images, based on the facial expressions of people, can add value in different areas such as education, shopping centers, hotels, entertainment centers, restaurants, among others, since it allows a better understanding of the requirements of the people, improve services, and predict sales trends. In a classroom, this technology allows to identify in real time the reaction of students to the development of the class, and in this way, the teacher can make the necessary adjustments to improve the learning process. The first step for this application is to detect faces of multiple students present in the scene, with efficient algorithms that process good-quality images. In this paper, the performance of six face-detection algorithms is determined using images taken in a classroom, in the town of Túquerres, in the department of Nariño, Colombia. The results show that a good camera resolution of 5 megapixels or higher, and good lighting conditions are determinant for successful face detection in classrooms of approximately 46 m2. In addition, the best performance was obtained with RetinaFace algorithm, which is more robust to different facial postures, achieving an accuracy of 96.5% with poor lighting conditions and 97.84% with good lighting conditions.
KW - Classroom
KW - Emotion Recognition
KW - Face Detection
KW - Feedback
KW - Group of Students
KW - Performance Evaluation
UR - http://www.scopus.com/inward/record.url?scp=85178617657&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-48642-5_2
DO - 10.1007/978-3-031-48642-5_2
M3 - Contribución a la conferencia
AN - SCOPUS:85178617657
SN - 9783031486418
T3 - Lecture Notes in Networks and Systems
SP - 14
EP - 20
BT - Proceedings of the 15th International Conference on Ubiquitous Computing and Ambient Intelligence (UCAmI 2023) - Volume 2
A2 - Bravo, José
A2 - Urzáiz, Gabriel
PB - Springer Science and Business Media Deutschland GmbH
T2 - 15th International Conference on Ubiquitous Computing and Ambient Intelligence, UCAmI 2023
Y2 - 28 November 2023 through 29 November 2023
ER -