Role of Deep Neural Features Vs Hand Crafted Features for Hand Written Digit Recognition
Jyostna Devi Bodapati1, B Suvarna2, Veeranjaneyulu N3
1Jyostna Devi Bodapati, Asssistant Professor, Department of CSE, Vignan’s University, Guntur (Andhra Pradesh), India.
2B Suvarna, Asssistant Professor, Department of CSE, Vignan’s University, Guntur (Andhra Pradesh), India.
3Veeranjaneyulu N, Professor, Department of IT, Vignan’s University, Guntur (Andhra Pradesh), India.
Manuscript received on 12 February 2019 | Revised Manuscript received on 02 March 2019 | Manuscript Published on 08 June 2019 | PP: 147-152 | Volume-7 Issue-5S4, February 2019 | Retrieval Number: E10280275S419/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Handwritten digit recognition can be considered as a subtask of hand written character recognition, a broad area where a given character is recognized automatically by a machine. The major challenges of hand written character recognition are: writing style and size of characters varies from person to person. With the advances in machine learning algorithms the success of handwritten character recognition is improved. In this task we have considered hand written digit recognition, as there are plenty of real-time applications like amount identification on Bank cheques, recognizing zip codes on postal letters to mention few. Recent literature shows that performance of Convolution Neural Network (CNN) is promising on images. We have used neural network based models for hand written digit classi-fication. Initially the model is trained on MNIST dataset. In this work we have tried to identify the effect of different types of features on the performance of the model.
Keywords: Deep Learning, CNN, Hand Crafted Features, Hand Written Character Recognition, Pooling, Convolution, Dropout.
Scope of the Article: Deep Learning