Zilla-64: A Bangla Handwritten Word Dataset Of 64 Districts Name of Bangladesh and Recognition Using Holistic Approach
Bangla Handwritten Word Recognition (BHWR) is a very challenging task due to the high curvature nature of the character, overlapping between characters, and flourishes in the writing style of Bangla Handwritten Word (BHW). Despite the importance and challenges of BHWR, the handwritten word dataset of Bangla is very few. In this paper, a new challenging Bangla word dataset which, we called ‘Zilla-64’, is introduced, and also the preparation of the dataset is discussed. To the best of our knowledge, this is the first well-labeled Bangla word dataset. This dataset can be used for gender, age, and education level handwritten related researches. Deep learning shows tremendous success in the handwritten recognition area. Because of the popularity of deep learning methods and for testing the performance of the dataset, a holistic approach based, Deep Convolutional Neural Network (DCNN) is applied on the dataset and achieved 93.30% accuracy.
If you want to cite our papers, please use:
@INPROCEEDINGS{9642594,
author={Azad, Md Ali and Singha, Hijam Sushil and Nahid, Md Mahadi Hasan},
booktitle={2021 International Conference on Science & Contemporary Technologies (ICSCT)},
title={Zilla-64: A Bangla Handwritten Word Dataset Of 64 Districts` Name of Bangladesh and Recognition Using Holistic Approach},
year={2021},
volume={},
number={},
pages={1-6},
keywords={Deep learning;Handwriting recognition;Urban areas;Education;Writing;Stability analysis;Robustness;Convolutional Neural Network;Holistic Approach;Bangla Handwritten Word;Bangla Handwritten Word Recognition;Zilla-64},
doi={10.1109/ICSCT53883.2021.9642594}}