Implementasi Convolutional Neural Network Untuk Deteksi Nyeri Bayi Melalui Citra Wajah Dengan YOLO

• Main Authors: Tomy Abuzairi, Nurdina Widanti, Arie Kusumaningrum, Yeni Rustina
• Format: Article
• Language: Indonesian
• Published: Ikatan Ahli Indormatika Indonesia 2021-08-01
• Series: Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi)
• Subjects: nvdia jetson nano developer kit, pain, baby, yolo, pytorch, cnn, facial expresion.
• Online Access: http://jurnal.iaii.or.id/index.php/RESTI/article/view/3184

Pain in a baby is difficult to detect is because the method for detecting pain is self-reporting even though babies themselves still cannot describe the pain verbally, then by observing changes in behavior in the form of facial expressions. Statistically, it is also recorded that about 80% of the world's population pays less attention to pain assessment, especially for children, even though this pain gives children a bad experience so that it can interfere with pain responses in the future or psychological trauma. Based on these problems, a prototype system was made using the NVIDIA Jetson Nano Developer kit to help detect pain, especially in infants 0-12 months by using the Convolutional Neural Network (CNN) model with the PyTorch framework and the You Only Look Once (YOLO) algorithm with three detection classification is sad, neutral and sick. From the results of the study, it was found that the YOLO algorithm was able to detect the three classifications with a sad mAP value of 77.8%, neutral 76.7%, in pain 68.9%. With a precision value of 71.4%, recall 62.5% and f1-score 66.6%. The average value of Confidence is 53.57%.