Automatic Multiple Choice Question Evaluation Using Tesseract OCR and YOLOv8

This paper presents a novel approach for automating the grading of multiple-choice question (MCQ) answer sheets using computer vision and pattern recognition techniques. The system examines student’s marked answer sheet images by comparing with the question sheet image and answer keys. The computer vision and pattern recognition helps extracting pertinent data such as question number detection, MCQ option detection and the answer markings. The proposed approach reliably produces the output report that displays the students’ correct answers with an accuracy of 0.98 F1 score and 0.99 mAP from any form or unstructured question script. This approach can provide a dependable and effective grading system, reducing manual work and offering prompt feedback to students without any constraints on the answer sheets.