Using Machine Learning and Image Processing to Effectively Analyze Scanned Answer Scripts after an Examination

The growth of artificial intelligence over the years has managed to solve some of the most difficult problems in the field of technology. The applications of machine learning are vast. This paper attempts to propose a way to use artificial intelligence to increase the efficiency of paper analysis. The aim of the paper is to use machine learning and image processing for the effective digital transformation of answer papers after an examination. In a large portion of schools in India, after every test correction, there are certain procedures that need to be followed. These procedures make the task cumbersome, time consuming and prone to errors. In order to tackle the problem, a program was developed to work on the scanned copies of answer scripts for generating an excel file consisting of the analyzed data.

First, the answer script was redesigned to increase the efficiency of the OpenCV algorithms. Secondly, the various functions of the OpenCV library were used to collect data from papers. Finally, the collected data was then passed through the MNIST dataset for recognition of handwritten digits and generation of excel file.

The program is able to identify the location of all the handwritten data within the answer script (100%) and is able to predict the handwritten digits up to an accuracy of 92%. In order to eliminate errors, an easy checking algorithm is also made. The application of artificial intelligence in paper analysis has many advantages ranging from reduced effort to time saving.