<![CDATA[This study aims to develop a model for detecting enthusiasm levels in online learning using the YOLOv11 algorithm, enhanced through hyperparameter optimization. Facial expressions serve as crucial indicators in determining enthusiasm, as they reflect the level of attention and interest a learner has toward the material. By increasing the number of interest level categories, the model is expected to provide a more detailed and accurate assessment of student engagement. The dataset used in this research is sourced from FER2013, which initially consists of seven emotion classes. These emotions are reorganized and classified into five enthusiasm levels to represent different levels of interest in learning better. Each level contains 1,000 images, resulting in a dataset of 5,000 images. This dataset was refined from previous studies to enhance its relevance and improve detection performance, making it more suitable for real-world applications. To achieve optimal performance, key hyperparameters, including the number of epochs, batch size, and image size, were fine-tuned. Before optimization, the model demonstrated an average precision (mAP 50-95) of 95.2% with an inference time of 1.7 milliseconds. After hyperparameter tuning, the model’s performance improved significantly, reaching an average precision (mAP 50-95) of 97%. However, this enhancement came with a slight increase in inference time to 3.1 milliseconds. The results highlight that fine-tuning model parameters can enhance detection accuracy while maintaining efficient processing speed, making it highly applicable in educational settings for assessing learner engagement.]]>
