Handwritten Document Conversion Using Regional Convolutional Neural Network
Loading...
Files
Date
2018-06-19T09:38:51Z
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This project converts handwritten document into its corresponding digital format using Regional Convolutional Neural Network (R-CNN). R-CNN is the pre-eminent machine
learning visual object detection algorithm. Since, digital documents can be easily manipulated, stored and retrieved, this project serves various applications such as processing cheques in banks, converting handwritten books into digital copies that can
used for publication, retrieving information from application forms, etc. Firstly, the handwritten document image is pre-processed to produce better results using adaptive thresholding, which removes noise and artifacts in the image. Secondly, character segmentation is performed in order to obtain individual handwritten characters from the image. The position of each of these characters is retrieved in order to position the
characters in the final text document. Thirdly, these characters are passed one-by-one
into the Convolutional Neural Network (CNN) model, which performs classification and produces the probability of each character class with one character class having the highest probability using softmax function. The character with the highest-class probability is chosen as the output character and is written into the text document based on its corresponding position.
Description
Keywords
MOHAN KRISHNA B, SWATHI MADHAVAN, PRAKASH H, RAHUL N, 1NH14CS071, 1NH14CS732, 1NH14CS167, 1NH14CS098