Hand Written Text Recognition
Loading...
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT,LAHORE
Abstract
The Handwritten Text Recognition project pertains to the current problem, conversion of
images containing written text into machine-readable formats, which can be modified,
using the power of deep learning. Until now, Optical Character Recognition (OCR) was
quite effective in most aspects, however recognition of handwritten stimuli still renders
primitive. This is due to the extreme variability of individual handwriting, the degree of
neatness and consistency of writing, and also the form of letters. This project sets out to
overcome these problems by building a model of long short term memory networks
(LSTM) and convolutional neural networks (CNNs) coupled with Connectionist
Temporal Classification (CTC) for the identification and interpretation of handwritten
text images.
Aim of this project is to construct and deploy accurate and efficient high volume text
recognition system, which would tolerate different handwriting styles. Input is a
handwritten text image which is passed through several layers of a neural network in
order to extract features and then those features are fed into CTC layer for generation of
text. The model has been trained on various entities of handwriting for the period of its
construction so as to make it effective on several hospitable writing styles and all their
variants.