handwriting digital recognition system using machine learning algorithms

• 2020-12-31 03:23:47
• (0) Comments
• 733 View

handwriting digital recognition system using machine learning algorithms

Recently handwritten digit recognition becomes vital scope and it is appealing to many researchers because of its use in a variety of machine learning and computer vision applications. However, there are deficient works accomplished on Arabic pattern digits because Arabic digits are more challenging than English patterns. Hence, the lacking research on using Arabic digits endeavors us to dig deeper by creating our challenge Arabic Handwritten Digits which consists of more than 45,000 samples. As a challenging dataset is used for evaluation, a robust deep convolutional neural network is used for classification and superior results are achieved.

Recently Deep Convolutional Neural Networks (CNNs) becomes one of the most appealing approaches and have been a crucial factor in a variety of recent successful and challenging machine learning applications such as challenge ImageNet [1, 2, 3, 4, 5,24], object detection [1, 6, 7], image segmentation [9,10], and face recognition [11, 12, 13]. Therefore, CNNs are considered our main model for our challenging tasks of image classification. Specifically, it is used for handwriting digit recognition which is one of the high academic and business transactions [14]. Handwriting digit recognition application is used in different tasks of our real-life purposes. Precisely, it is used in banks for reading checks, post offices for sorting letters, and many other related tasks.

Apparently, English Handwriting datasets are widely available, and significant achievements have been made for English digit datasets such as CENPARMI [15], CEDAR[16], and MNIST[17], however, there are rare works accomplished on Arabic digit datasets for many reasons. One of the critical factors that can influence working on the Arabic dataset is lacking to the dataset. The unavailability of datasets can be one of the essential factors that can diminish working on Arabic datasets. Hence, the deficiency of a large challenging Arabic dataset strives us to extensively working on creating the largest and most challenging dataset which contains more than 45,000 patterns. Furthermore, we investigate and demonstrate a powerful DCNN used for classification. Not only designing a powerful DCNN is presented but also critical parameters of CNN is carefully selected and tuned to produce a final concrete model which achieves superior results.

MODULES:

1. Preparing patterns before feeding to CNN. All images are pre-processed before passing into the network. In our experiments, CNN is designed to receive an image size of 64x64 pixels. Hence, all images have been cropped to the same size to be fed to the model.
2. After preparing images, they are fed to the deep model to extract features. As demonstrated earlier a robust CNN is used in this experiment to extract robust features used in the final decision to justify the class to which they belong.
3. Finally, the last layer named the softmax layer is used at the top of CNN to minimize the error.

Algorithm used

CNN

supported operating system

• 1. window
• 2. Linux
• 3. ubuntu

Cost and technology

CONCLUSION

In this work, a new challenging digit Arabic dataset is collected from different study levels of schools. A large dataset is collected after paying vast effort for distributing and collecting digit forms over hundreds of primary, high, college students. After we find that there are few and not challenging Arabic digit datasets, we paid vast effort into preparing such a challenging dataset. Also, the collected dataset is trained using an efficient model of CNN which represents the current state-of-the-art for a variety of applications. Thus we extensively analyzed the model by carefully selecting its parameters and showing its robustness for handling our dataset.

Project demo