An Automated System to Calculate Marks from Answer Scripts
Redwan Ahmed Rizvee, Md. Fahim Arefin, MD. Rayhun Khan, MD. Nezarul Islam, Kh. Fazle Rabbi · 2022
During evaluating exam scripts, teachers have to write down marks for each question on the top sheet of the answer script and then sum those marks to calculate the total marks - which is a common source of error. This literature has focused on automating the script marks calculation process using multiple handwritten characters' recognition. Traditional solutions for identifying handwritten characters focus on labeling a single character using suggestion based solution. So, existing solutions are not enough to address the problem that this study has focused on. In this regard, the proposed framework has tried to develop a hybrid approach. The proposed solution first identifies the connected components, based on the pixel color as red, through a depth-first search based solution. Then each connected component is passed through a learning architecture to generate a numeric digit suggestion combining heuristic and machine learning, making a hybrid system. The proposed system is very modular and flexible to be implemented and modified. The extensive analysis presented in the evaluation section shows the merit of the proposed framework in automating the addressed problem.