Real-time classroom behavior and drawing pattern analysis using YOLOv11 for multi-class student activity detection

Students are important for the growth of a country. Student attention detection plays an important role in ensuring effective learning in the classroom. It is an important area of research as it has the potential to improve the quality of education and help maintain integrity during exams. Further, the students’ monitoring is not only helpful in exams, but it is also useful for educators in maintaining discipline in class, such as it helps in the identification of misconduct and allows timely intervention. With the advancement of technology, researchers proposed several methods for students’ classroom behavior classification that rely on traditional feature extraction algorithms and machine learning classifiers. These methods faced several challenges, such as the complexity of features, and the presence of more than one class in the same image, which may lead to false classification results. With the emergence of deep learning, recent research studies have focused on the CNN models, but there is still a gap in the domain. In this research work, we have proposed a method based on the advanced deep learning model YOLOv11 for students’ behavior detection. The proposed method consists of several major phases, such as preprocessing, model fine-tuning, hyper-parameter setting, Model training, and evaluation of results. The study utilized three publicly available datasets to train the model for analyzing the student class behavior, class attention, and observing the behavioral patterns from the drawing of students. The datasets are preprocessed to improve the contrast, which is later fed to the fine-tuned YOLOv11 model for training. The proposed method achieved the highest mAP of 96.8%. The results of the model show that it has the potential to be implemented in a real-time classroom monitoring environment. Code: https://www.kaggle.com/code/jianaz786/v5-students-behavior-yolo57811