Automated Examination Grading Using Deep Learning Categorization Techniques

EasyChair Preprint 857

Abstract

In previous time, students whatever from junior or senior have lots of examination in school. As the competition for learning grows, the fre- quency of exams increases. It takes time for teachers to mark these exam papers, students also need to wait for a long time to get their score. This paper is focus on estimating the score by using deep learning methods. This method of scoring is based solely on the image characteristics of the answer sheet, rather than scoring through text-based methods after recog- nizing the text. Experimental results shows that We are able to achieve an accuracy of approximately 85% over a 10 scores error and approximately 95% accuracy over a 20 scores error. And we can use the deep convo- lutional neural network we trained as the discriminator of Generative Adversarial Network to generate high score papers and improve student achievement.

Keyphrases: Classification, computer vision, deep learning, machine learning, recognition

@booklet{EasyChair:857,
  author    = {Lehan Yang},
  title     = {Automated Examination Grading Using Deep Learning Categorization Techniques},
  howpublished = {EasyChair Preprint 857},
  year      = {EasyChair, 2019}}