This website requires JavaScript.

Huruf: An Application for Arabic Handwritten Character Recognition Using Deep Learning

Minhaz KamalFairuz ShaiaraChowdhury Mohammad AbdullahSabbir AhmedTasnim AhmedMd. Hasanul Kabir
Dec 2022
摘要
Handwriting Recognition has been a field of great interest in the ArtificialIntelligence domain. Due to its broad use cases in real life, research has beenconducted widely on it. Prominent work has been done in this field focusingmainly on Latin characters. However, the domain of Arabic handwritten characterrecognition is still relatively unexplored. The inherent cursive nature of theArabic characters and variations in writing styles across individuals makes thetask even more challenging. We identified some probable reasons behind this andproposed a lightweight Convolutional Neural Network-based architecture forrecognizing Arabic characters and digits. The proposed pipeline consists of atotal of 18 layers containing four layers each for convolution, pooling, batchnormalization, dropout, and finally one Global average pooling and a Denselayer. Furthermore, we thoroughly investigated the different choices ofhyperparameters such as the choice of the optimizer, kernel initializer,activation function, etc. Evaluating the proposed architecture on the publiclyavailable 'Arabic Handwritten Character Dataset (AHCD)' and 'Modified Arabichandwritten digits Database (MadBase)' datasets, the proposed modelrespectively achieved an accuracy of 96.93% and 99.35% which is comparable tothe state-of-the-art and makes it a suitable solution for real-life end-levelapplications.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答