A YOLO-based software framework for automated paper-based multiple-choice exam scoring

This paper presents an open-source software framework for automated scoring of paper-based multiple-choice exams. The proposed framework processes scanned or photographed answer sheets, then corrects distortions, extracts student data, and recognizes selections with high accuracy. The framework can also generate structured outputs and annotated images for easy integration into learning platforms. Evaluations on real university exams show that the framework achieves 50 sheets per minute with the error rate below 0.5% on commodity hardware, thereby enabling cost-effective assessment at scale.