This report is submitted as partial fulfillment of the requirements of the Computer Arabizition course’s practical exam
Our project's aim is to program the computer to identify hand-written alphabets via matrix operations. Each alphabet image contains 32*32 pixels, and we create a matrix using these pixels. By multiply the matrix to several sample matrixes, the pixels are converted into a Deep neural network. And finally, we employ adam optimizer method so that the computer can predict the highest possibility of the alphabet written.
Keywords: deep learning, ANN, Feature Extraction, CNN, Arabic, Machine Recognition, natural and physical sciences, image data, image processing
The main objective of this research is to find a new solution for handwritten text recognition of different fonts and styles by improving the design structure of the traditional Artificial Neural Network (ANN). ANNs have been successfully applied to pattern recognition, association and classification, forecast studies, and control applications, to name a few. The recognition results of such text or handwritten materials are then fed into Optical Character Recognition (OCR) as an electronic translation of images of handwritten, typewritten or printed text into machine-editable text. OCR is a field of research that is fully developed and has been quite useful in pattern recognition, artificial intelligence and machine vision. Consequently, typewritten text recognition that is void of any distortions is now considered largely a solved problem. However, the direct use of OCR on handwritten characters remains a very difficult problem to resolve, yielding extremely low reading accuracy. handwritten document recognition is currently a difficult problem; as different people have different handwriting styles. Scanning, segmentation and classification are the general processes that are being used to recognize handwritten documents. ANNs have proven to be excellent recognizers of printed characters and handwritten characters.
Character recognition is one of the most important research fields of image processing and pattern recognition. Character recognition is generally known as Optical Character Recognition (OCR).OCR is the process of electronic translation of handwritten images or typewritten text into machine editable text. It becomes very difficult if there are lots of paper based information on companies and offices. Because they want to manage a huge volume of documents and records. Computers can work much faster and more efficiently than human. It is used to perform many of the tasks required for efficient document and content management. But computer knows only alphanumeric characters as ASCII code. So computer cannot distinguish character or a word from a scanned image. In order to use the computer for document management, it is required to retrieve alphanumeric information from a scanned image. There are so many methods which are currently used for OCR and are based on different languages. The existing method like Artificial Neural Network (ANN) based on English Handwritten character recognition needs the features to be extracted and also the performance level is low. So a Convolutional Neural Network (CNN) based English handwritten character recognition method is used. It's a deep machine learning method for which it doesn't want to extract the features and also a fast method for character recognition.
My personal motivation is that I Have faced a problem, where I wanted to transform a paper document into a digital one, but I had to type it character by character. I have been trying to find an easy way to solve the problem,simply The purpose of our project is to recognize hand-written alphabets, so the computer can automatically identify the characters without any manual input. The link to my datasource is:here
This project requires Python and the following Python libraries installed:
You will also need to have software installed to run and execute a Jupyter Notebook
If you do not have Python installed yet, it is highly recommended that you install the Anaconda distribution of Python, which already has the above packages and more included.
code is provided in the Arabic Handwritten Character recognizer.ipynb notebook file.
In a terminal or command window, navigate to the top-level project directory Arabic-Chracters-OCR/ (that contains this README) and run one of the following commands:
ipython notebook Arabic Handwritten Character recognizer.ipynbor
jupyter notebook Arabic Handwritten Character recognizer.ipynbThis will open the Jupyter Notebook software and project file in your browser.
The A_Z Handwritten dataset contains containing handwritten images in size 3232 pixels, each alphabet in the image is centre fitted to 2020 pixel box.Each image is stored as Gray-level. The images manily are taken from NIST(https://www.nist.gov/srd/nist-special-database-19)
you can find the dataset here (https://www.kaggle.com/mloey1/ahcd1/kernels)
A Convolutional Neural Network (CNN) is a special type of feed-forward multilayer trained in supervised mode. The CNN trained and tested our database that contains 16798 of handwritten english characters.
Generating a confusion matrix,for summarizing the performance of a classification algorithm. Classification accuracy alone can be misleading if you have an unequal number of observations in each class or if you have more than two classes in your dataset. Calculating a confusion matrix can give you a better idea of what your classification model is getting right and what types of errors it is making.