I In the classroom learning process, the emotional condition of students plays an important role in determining their attention, engagement, and the overall effectiveness of learning activities. however, in practice, educators often face limitations in monitoring students\\\' emotional conditions continuously and comprehensively. therefore, the use of computer vision technology in a smart classroom system becomes a relevant solution to support a more adaptive learning environment. a facial emotion classification system enables the monitoring of students’ affective conditions through the analysis of facial expressions captured by a camera. this research focuses on comparing the accuracy performance of the yolov11 and yolov26 algorithms in performing facial emotion classification in a smart classroom environment. the research method includes the stages of facial image data collection and preprocessing, model training using both yolo algorithm variants, and system performance evaluation based on accuracy and inference speed metrics. the system implementation is carried out in a smart classroom scenario to represent real-world classroom conditions. the expected result of this research is to present an objective performance comparison between yolov11 and yolov26 models in facial emotion classification. this comparison is intended to serve as a foundation for determining the most efficient and optimal detection algorithm for the development of emotion recognition-based smart learning systems.