Handwritten Digit Recognition Using CNN

EasyChair Preprint 9618

Abstract

In this paper, we are going to see how we can train a neural network model to recognize a handwritten digit which is given as an input to the model. The algorithm used to realize it is Convolution Neural Network (CNN).It is a network architecture for deep learning. It learns from the data which isthrough the images. It finds patterns in images and recognizes objects and categories. The CNN has several layers which takes the input, analyzes the input, and producesoutput. It's very much used in Deep Learning and very efficient in the modern world filled with AI. It's a part of ANN that has been the superior algorithm incomputer vision tasks. It has achieved top-level performances in various fields like medical research, AI, etc. CNN, which is used for processing data, is a type ofdeep learning model that has grid patternsi.e.-images. It is a construct that has three kinds of layers namely convolution, pooling, and fully connected layersrespectively.The first few layers do the feature extraction, whereas the next layer maps the extracted features into final output. The convolution layer plays an importantrole in CNN. It is composed of a stack of mathematical operations such as convolution. The pixel values are stored in a twodimensional array in the digital images and small grid of parameters called kernel is applied at each image position. This makes the CNN highly efficient for image processing. The layers perform convolutionand subsampling one after another. Output of one layer is input of the next layer. The output of the final layer is our predicted value. Extracted features can progressivelybecome more complex, as one layer feeds its output to the next layer.

Keyphrases: Handwriting Recognition, Relu function, character recognition, deep learning, feature extraction

@booklet{EasyChair:9618,
  author    = {V Viswanatha and A C Ramachandra and Satya Dev Nalluri and Sai Manoj Thota and Aishwarya Thota},
  title     = {Handwritten Digit Recognition Using CNN},
  howpublished = {EasyChair Preprint 9618},
  year      = {EasyChair, 2023}}