Automated Inputting of Grades into Online Gradebook Using Optical Character Recognition

Abstract

For teachers with large classes, inputting grades is time consuming and error prone. We propose a more efficient and accurate process based on an optical character recognition (OCR) tool. The teacher takes two pictures, one of the name, and another of the grade, that are recognized and displayed for verification and manual correction, if needed. Then, the confirmed data will be recorded in the gradebook. We tested that the proposed process is faster and easier than manual input. To train the OCR with numeric data and name data, we used the machine learning library Scikit-learn with selected datasets. We achieved 86% accuracy with reading double digits and 62% accuracy with reading names. To test the recording time and accuracy, we fed 20 new self-produced images of grades and names into each model. If the name was not recognized in the grading book, we selected the name with the shortest distance, which improved the accuracy for names. 75% of the grades were assessed correctly, but only 5% of names were assessed correctly. The time it took to execute name and grade processing was 300 seconds. In comparison, manually grading resulted in only one wrong grade, and was completed in 158 seconds. While the results of the OCR models lack in accuracy and time efficiency, they could improve over time if used in actual classroom settings, as they would have access to real-word images of handwritten names as well as more images of handwritten names and grades in general.