YOLO-Based Object Detection for Locating Handwritten X-Marks on Multiple-Choice Answer Sheets

Abstract

Student assessment is considered a measure of learning achievement. Among the various assessment tools available, multiple-choice tests are commonly used, allowing students to mark their answers by filling or marking crosses on provided answer sheets. However, this method presents significant challenges, particularly when dealing with a large number of questions, as it may require additional personnel for grading, leading to delays in the assessment process. Additionally, manual grading can introduce errors, resulting in inaccuracies in score calculation. This study aims to address these challenges by developing a mathematical model capable of automatically detecting crosses on multiple-choice answer sheets using object detection techniques and deep learning. The model is based on the You Only Look Once version 8 (YOLOv8) technique for object detection. It employs a Convolutional Neural Network (CNN) for cross detection. The performance of the model is evaluated using metrics such as Precision, Recall, and F1-score. The results indicate that at a Threshold value of 0.496, the model achieves the highest F1-score of 0.989. This suggests that the model is effective in automatically detecting and identifying cross markings on multiple-choice answer sheets.