Using Artificial Intelligence To Grade Student's Physicial Paper Asessments

The Project

Due to the ongoing pandemic and study from home situation, students are spending a lot more time on screens than usual. This is causing screen fatigue and eye strain to many students, including me. My project is an attempt to reduce the amount of time students spend on screens to complete assessments and tests. For this project I have built a deep learning model to automatically grade physical paper assessments with multiple choice questions. My project has two parts. First, train a deep learning model that will make predictions on what multiple choice answer was shaded. Then, use that model and make a grading system to grade quizzes and assessments and display the score of each paper. My final model had an accuracy, loss, validation accuracy, and validation loss of 0.24, 0.93, 0.12, 0.96 respectively. The grading system scans the image of the physical assessment, predicts the answers shaded by the student for each question and compares them with the answer keys for the assessment and finally calculates the score and prints the score on the image. If I had more time for this project, I would try my best to improve my model’s accuracy. I have already increased the accuracy by adding color to my pictures while training my model. That helped a lot, so I would try to find methods like that to improve my model even more At the beginning of this project, my code was based on code from all the courses and tutorials I took to learn the subjects needed for this project. But as I got further into this project, the code I needed became more and more specific, forcing me to write a lot of my own code that I formed from countless hours of researching. The deep learning model developed for this project can also be used post pandemic to automate physical paper assessment grading in schools. It will help teachers cut down massively on the time they spend on grading assignments by letting them just take pictures of the assessments and automatically grade them in a split second.

View this project on YouTube