An Efficient off line Hand Written Character Recognition using CNN and Xgboost
Joseph JamesS1, C. Lakshmi2, Uday Kiran P3, Parthibano4
1Joseph James S, Department of Software Engineering, SRMIST, Kancheepuram (Tamil Nadu), India.
2Dr. C. Lakshmi, Department of Software Engineering, SRMIST, Kancheepuram (Tamil Nadu), India.
3Uday Kiran P, Department of Software Engineering, SRMIST, Kancheepuram (Tamil Nadu), India.
4Parthiban, Department of Software Engineering, SRMIST, Kancheepuram (Tamil Nadu), India.
Manuscript received on 07 April 2019 | Revised Manuscript received on 20 April 2019 | Manuscript published on 30 April 2019 | PP: 115-118 | Volume-8 Issue-6, April 2019 | Retrieval Number: F3408048619/19©BEIESP
Open Access | Ethics and 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: The purpose of this paper is to legitimize and implement the usage of Convolutional neural networks (CNN) in parallel with XGBoost model to improve handwriting Recognition systems. The usage of CNNs in recognizing handwritten characters is a broadly researched project yet the inclusion of different types of classification models along with CNN is sparse. The learning model proposed in this paper is based on (CNN) as a feature extraction tool and XGBoost as an accurate prediction model. The XGBoost gradient boosting model is evaluated for loss function and regularization and an appropriate objective function is decided. With the proposed method in which CNN and XGBoost are used together there is an expected increase in accuracy rate and total computation time. The model is trained and evaluated using the NIST special database 19 dataset which consists of 810,000 isolated character images including lower case, upper case and digits in the english language. The improvement in accuracy is in comparison with the handwriting recognition model which uses CNN alone and is augmented with the use of tree ensembles model which is XGBoost. The improved accuracy percentages are specified separately for lowercase letters, uppercase letters and numeral characters.
Keyword: XGboost-Extreme Gradient Boosting, CNN Convolutional Neural Networks, NIST- National Institution of Standards and Technology.
Scope of the Article: Image Processing and Pattern Recognition