Devanagari Digit and Character Recognition Using Convolutional Neural Network

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 02 | Sep 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 313
Devanagari Digit and Character Recognition Using Convolutional
Neural Network
Suresh Mestry1, Bhagyasha Patil2, Anushka Pede3, Shreya Raul4
1 Assistant Professor, Department of Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology,
Maharashtra, India
2 Student, Department of Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Maharashtra, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - In recentdecades, ConvolutionalNeuralNetwork
(CNN) has achieved remarkable results in both the research
field and the application field due to the significant
achievement acquired in image recognition which is
abundantly used in the field of machine learning. Handwriting
recognition is the bridge between handwriting and machines,
and can play a huge role in finance, accounting, educationand
other fields reducing labor cost. At present, the recognition of
handwriting has received greatdegreeofattentionfrommany
researchers. Handwritten digit and character recognition
becomes vital scope and it is appealing many researchers
because of it’s using in variety of machine learning and
computer vision applications. However, due to its complexity
handwritten digit and character recognition still has a great
development space. The recognition of handwritten digitsand
character remains a difficult problem, particularly in some
scripts, there exists a large variation of style across writers.
This large variation is an interesting challenge for algorithms
in image processing and pattern recognition. There are
deficient works accomplished on Devanagari scripts because
Devanagari scripts are more challenging than English. The
proposed system evaluateshandwrittenDevanagaricharacter
and digit recognition. CNN model is used to determine the
result to achieve character and digit recognition. This system
will be implemented using Python and its different library
functions.
Key Words: CNN, Handwritten, Machine Learning,
Devanagari.
1. INTRODUCTION
Over the period of time the numbers of fields in which deep
learning is applied is increasing. In deep learning, visual
imagery analysis uses Convolutional Neural Networking
(CNN). Object detection, video analysis, segmentation,
pattern recognition, face recognition, natural language
processing, spam detection, topic categorization, regression
analysis, speech recognition, imageclassificationaresomeof
the examples that can be done using Convolutional Neural
Networks. The accuracies in various such fields including
handwritten digits recognition using Deep Convolutional
Neural Networks has reached human level perfection.
Devanagari script is based on the ancient Bramhiscriptused
in the Indian subcontinent. Devanagari Script consists 47
primary characters including 14 vowels and 33 consonants.
Handwritten character and digit recognition refers to the
automatic identification of handwritten alphabets and
numbers through computers or other equipment.
2. OBJECTIVES
 To overcome the difficulty in recognitionofDevanagari
digit and characters due to variation in writing patters;
using Convolutional Neural Network.
 Perform an efficient pre-processing to make data
suitable for analysis.
 Create a CNN model to perform classification process
by training the model with training data to fit and test
the model to evaluate the accuracy rate.
3. PROPOSED SYSTEM
We intend to develop a system where the application
recognizes Devanagari digits and characters based on CNN.
In the recognition of handwritten characters and digits
particularly in the Devanagari script; there exists a larger
variation of style across writers which is an interesting
challenge in image pre-processing and recognition.
4. METHODOLOGY
A Convolutional Neural Network, which is also known as
CNN or ConvNet, is a class of neural networks that
specializes in processing data whichhasa grid-liketopology,
such as an image. A digital image is a binary representation
of visual data which contains a series of pixels arranged in a
grid-like fashion that contains pixel values to denote the
brightness of the pixel and what color each pixel should be.

Fig -1: Digital image with pixel values
Layer helps a Neural Network (NN)togaincomplexityin any
problem. Increasing number of layers (with units) can
increase the non-linearity of the output of an NN. Each layer
contains some amounts of Units. The amount is entirely up
to the creator. However, having too many layers for a simple
task can unnecessarily increase its complexity and decrease
its accuracy. The opposite also holds true. Every NN has 2
layers. Those are the input and output layers. Layers in
between those are called hidden layers. The NN in the
picture below contains an input layer (with 8 units), an
output layer (with 4 units) and 3 hidden layers with each
containing 9 units.
Fig -2: Deep Neural Network
A neural network with 2 or more hidden layers with each
layer containing a large amount of units is called a Deep
Neural Network.
A CNN typically has three layers:
 Convolutional layer
 Pooling layer
 Fully connected layer.
Fig -3: CNN hidden layers
The core building block of the CNN is convolution layer. It
carries the main part of the network’s computational load.It
performs a dot product between two matrices, where first
matrix is the set of learnable parameters known as a kernel,
and the second matrix is the restricted portion of the
receptive field. The kernel is smaller than an image but has
larger depth. Which means, the height and width of the
kernel will be spatially small, but the depth will extend up to
all the three channels if the image is composed of three
(RGB) channels. During the forward pass, the kernel slides
across the height and width of the image whichproducesthe
image representation of that receptiveregion.Thisproduces
a two-dimensional representation of the image which is
known as an activation map which gives the response of the
kernel at each spatial position of the image.
Fig -4: Two- Dimensional Represenation
The pooling layer is used to replace the output of the
network at certain locations by deriving a brief statistic of
the nearby outputs. It helps in reducing thespatial sizeofthe
representation, decreasing the required amount of
computation and weights. The pooling operation is
individually processed on each and every slice of the
representation. There are various pooling functions such as
the average of the rectangle neighborhood, L2 norm of the
rectangular neighborhood and weighted average based on
the distance from the central pixel. Max pooling reports the
maximum output from the neighborhood and is the most
popular process.
Fig -5: Max Pooling
Neurons in this layer are fully connected with all neurons in
the preceding and succeeding layer as seen in regular FCNN.
This is why it can be computed by a matrix multiplication
followed by a bias effect. The FC layer assist in mapping the
representation between the input and the output.

Fig -6: FC Layer
5. ARCHITECTURE
Fig -7: The System Architecture
Figure below gives the system architecture of proposed
system. To train the model first the dataset is pre-processed.
After pre-processing feature extraction is done for the
dataset from which we get training images. These training
images are passed to the CNN model whichfurthergives usa
trained model. On the basis of this trained model
classification is done and output is generated. While testing
the model first the dataset is pre-processed. After pre-
processing feature extraction is done for the dataset from
which we get testing image. This testing image is passed on
to trained model and using CNN model classification is done
for the testing image and further output is generated. Now
when we give a real-time input, this input is passed to the
CNN model and using the trained model input is classified
and an appropriate output is generated.
Fig -8: System’s Detailed Design
6. EXPERIMENTAL RESULT
Our Kaggle dataset consists ofmultipleimagesofDevanagari
digits and characters. This dataset is used to train our CNN
model. The model trained with this dataset isusedtopredict
the output or we can say recognize the user input. The user
input is given as drawing the digit or character that are to be
recognized with the help of cursor on the whiteboard which
appears on the user interface. In total there are 92,000
images of Devanagari digits and characters. There are 10
digits from 0 to 9 and 36 characters from Ka to Nya each
having 2000 images respectively.
Fig -9: Dataset images
The above few images are from the dataset of Devanagari
character Ka and Devanagari digit 2 displaying variety in
handwritten patterns. The following images representshow
user input is to be given on the whiteboard window using
the cursor.

Fig -10: Devanagari Digit Recognition
Fig -11: Devanagari Character Recognition
Fig -12: Devanagari Digit and Character Recognition
7. CONCLUSIONS
Our model is based on multilevel classification using CNN is
useful in recognizingDevanagaridigitandcharacterswritten
in various styles. Using the classification obtained,
researchers can efficiently study the ancient Indian
literature.
ACKNOWLEDGEMENT
This system is implemented under the guidance of Prof.
Suresh Mestry, Department of Computer Engineering at
MCT’S Rajiv Gandhi Institute of Technology, Mumbai, India.
REFERENCES
[1] Burel, G., Pottier, I., &Catros, J. Y. (1992, June).
Recognition of handwritten digits by image processing
and neural network. In Neural Networks, 1992. IJCNN,
International Joint Conference on (Vol. 3, pp. 666-671)
IEEE.
[2] Niu, X.X.; Suen, C.Y. A novel hybrid CNN–SVM classifier
for recognizing handwritten digits. Pattern Recognit.
2012, 45, 1318– 1325.
[3] Wells, Lee & Chen, Shengfeng&Almamlook, Rabia&Gu,
Yuwen. (2018). Offline Handwritten Digits Recognition
Using Machine learning.
[4] Y. LeCun, B. Boser, J. Denker, D. Henderson, R. Howard,
W. Hubbard, and L. Jackel, "Handwritten digit
recognition with a back-propagation network," in
Advances in Neural lnformation Processing Systems
(NIPS) 2, 1990, pp. 396-404.
[5] D. Ghosh, T. Dube, and A. Shivaprasad, "Script
recognition: A review," IEEE Trans.PatternAnalysisand
Machine Intelligence, vol. 32, no.12,pp.2142-2161,Dec.
2010.
[6] Y. LeCun et al., "Backpropagationappliedtohandwritten
zip code recognition," Neural computation, vol. 1, no. 4,
pp. 541-551, 1989.
[7] L. Bottou, “Notes on convolutional neural networks”,
2006.
[8] Sueiras, J.; Ruiz, V.; Sanchez, A.; Velez, J.F. Offline
continuous handwriting recognition using sequence to
sequence neural networks. Neurocomputing.2018, 289,
119-128.
[9] Hubert Cecotti,“HandwrittenDigitRecognitionofIndian
Scripts”, IEEE, 2015
[10] Jinze Li, Gongbo Sun, Leive Yi, Qian Cao, Fusen Liang, Yu
Sun, “Handwritten Digit Recognition System based on
Convolutional Neural Network”, IEEE, 2020
[11] Mayank Jain, Gagandeep Kaur, Muhammad Parvez
Quamar, Harshit Gupta, “Handwritten Digit Recognition
using CNN”, IEEE, 2021

Devanagari Digit and Character Recognition Using Convolutional Neural Network

Recommended

More Related Content

Similar to Devanagari Digit and Character Recognition Using Convolutional Neural Network (20)

More from IRJET Journal (20)

Recently uploaded (20)

Devanagari Digit and Character Recognition Using Convolutional Neural Network