Applying Deep Learning and Kmeans Algorithms for Automated Multiple-Choice Exam Scoring

In recent years, the application of deep learning in automated scoring systems for multiple-choice exams has garnered significant attention. This paper presents a comprehensive approach to developing an automated scoring system that leverages the capabilities of deep learning models, particularly the YOLOv8 model and Kmeans algorithm, to accurately detect and score multiple-choice answer sheets. Our system processes scanned images of answer sheets through several key stages, including image preprocessing, template matching, and object detection using YOLOv8. We introduce a novel methodology to accurately identify student information and marked answers, utilizing the Kmeans algorithm for clustering detected answer bubbles. The dataset comprises both student information and answer sheets, and extensive preprocessing steps ensure uniformity and high-quality inputs for the model. The YOLOv8 model is employed to identify and locate the answer bubbles, achieving high accuracy and efficiency. Results demonstrate the system's capability to handle various answer sheet formats with minimal errors, significantly reducing the time and potential inaccuracies associated with manual grading. This research not only highlights the effectiveness of deep learning in educational contexts but also paves the way for future improvements and applications in automated scoring systems.