Handwriting Analysis Using Convolutional Neural Networks

  • Abstract
  • Keywords
  • References
  • PDF
  • Abstract

    Convolution is the technique to blend or overlap two or more functions. This technique when provided to artificial neural networks, works together to learn the features of different categories of objects and detects them based on its features instead of the shape and edges. This helps to detect the objects even in unusual positions. Since, features of an object remains constant, CNN provides high efficiency significantly better than traditional cascade methods. CNN networks follow convolution, max pooling, flattening. These process combines preprocess the image for training and then the image is transferred to artificial neural networks.



  • Keywords

    Convolution; Detection; Flattening; Identification

  • References

      [1] Jianxin Wu, 2017, Introduction to Convolutional Neural Networks

      [2] C.-C. Jay Kuo, 2016, Understanding Convolutional Neural Networks with A Mathematical Model

      [3] Dominik Scherer et al., 2010, Evaluation of Pooling Operations in Convolu-tional Architectures for Object Recognition

      [4] www.superdatascience.com/deep-learning

      [5] Monika Jain et al./ Elixir Digital Processing 88 (2015) 36377-36380 “Development of image processing software for online measurements at streak camera system in Indus-1 synchrotron radiation source”.

      [6] Upadhyay, J., Garg, Akash Deep, Ojha, Avanish, Tyagi, Y., Sharma, M.L., Puntambekar, T.A., Navathe, C.P., Vora, H.S., & Jain, Monika (2015). Measurement of longitudinal electron beam parameters using indigenously developed streak camera system at Indus-1 synchrotron radiation source. India: Bhabha Atomic Research Centre.

      [7] Monika Jain, Rahul Saxena “Parallelization of Video Summarization over Multi-Core Processors” February 2018, International Journal of Pure and Applied Mathematics 118(9)

      [8] http://yann.lecun.com/exdb/mnist/




Article ID: 24305
DOI: 10.14419/ijet.v7i4.41.24305

Copyright © 2012-2015 Science Publishing Corporation Inc. All rights reserved.