Identifikasi Gerakan Sholat Wajib dalam Menentukan Jumlah Rakaat Menggunakan Model MoveNet secara Real-Time

Abstract

Sholat is a form of worship performed by Muslims as an expression of devotion to God, requiring a high level of focus and concentration to be accepted by Allah SWT. A lack of concentration during sholat often leads to forgetting the number of rak'aahs or the sequence of sholat movements, especially when performed individually. Identification of sholat movements using camera can help to remind when errors occur during the execution of sholat postures. MoveNet offers a machine learning model capable of detecting human body positions during various activities. This model, built on a CNN backbone, provides lightweight computation, making it suitable for real-time applications that do not require large storage capacity while still maintaining high accuracy. Model training using keypoints extracted from MoveNet focusing on the movements of standing, sitting, bowing (ruku), and prostrating (sujud) during sholat achieved 97% accuracy across 100 tests using images captured at varying camera positions and distances.