Analysis of Deep Learning Accuracy for BISINDO Sign Language Translation

Rizal Sengkey (Dept. of Electrical Engineering, Sam Ratulangi University Manado)
Brave A. Sugiarso (Dept. of Electrical Engineering, Sam Ratulangi University Manado)
Dirko G. S. Ruindungan (Dept. of Electrical Engineering, Sam Ratulangi University Manado)

Ensiklopedia of Journal
Vol 8, No 10 (2026): Vol. 8 No. 10 Juli 2026

Communication between the deaf and hearing people is still a challenge, especially due to the lack of understanding of the sign language they often used. To overcome this, the author developed an Indonesian Sign Language (BISINDO) translation and recognition application by utilizing the YOLOv4 deep learning model. This study aims to analyze the accuracy of the model in detecting and translating sign language in real-time. The research stage begins with creating a model. The model will be trained using the BISINDO dataset which can be accessed through an open-source dataset provider site for computer vision, and the performance evaluation will be carried out through several metrics such as Intersection over Union (IoU), Precision, Recall, and Mean Average Precision (mAP). The expected result is a model that is able to recognize and classify sign language with high accuracy based on the predetermined metrics. This research is useful in providing inclusive technological solutions to facilitate communication between deaf and hearing people, while increasing accessibility in the use of sign language.Key words — Accuracy; BISINDO; Deep learning; Sign language.