Date

6-2017

Degree Name

BS in Computer Engineering

Department

Computer Engineering Department

Advisor(s)

Dennis Sun

Abstract

The objective of this project was to automate the process of grading handwritten numerical answers in a classroom setting. The final program accepts a scanned answer sheet completed by the student along with a description of the correct answers and produces a detailed report describing the confidence of correctness for each answer.

Computer vision techniques are used to automatically locate the locations of the answers in the scan. Each digit is then passed through a convolutional neural network to predict what was written by the student. The individual probabilities of each digit produced by the network are aggregated into a single score describing the model’s confidence in the correctness of the answer.

Share

COinS