Yolov10-Based Student Activity Identification as Personalization of Appropriate Learning Methods

Suitable learning methods can significantly improve the quality and learning outcomes of students. An independent curriculum personalizes learning methods to create an inclusive and student centred learning environment. This approach can be done through the Student-Centered Learning (SCL) method. However, most of the literature states that these efforts are only done by exploration using manual methods. That shows that there are limitations in the approach used so far. Therefore, this research focuses on the automatic detection of student learning activity engagement in the classroom. We implemented the You Only Looks Once version 10 Medium (YOLOv10-M) algorithm that has been trained using a specific dataset. Based on the training results, the model can recognize visual behaviors based on student engagement during classroom learning activities. The dataset used is 8,884 total images with 8 classes. The experimental results of the model produced excellent detection performance with an average precision (mAP@50) of 91.1%. The model is proven to have the potential to help teachers recognize students' learning activities in real-time in classrooms as an evaluation of learning methods and constructional adjustments.