The purpose of this project is to extract and recognize Persian Handwritten Digits/Alphabet from a form. this project has been developed as final project of Foundations of Computer Vision Course.
the accuracy of this work for digit recognition is about 98% and for letter recognition is about 96%.
##how to use:
the input is a form like the ones below (photo can be taken from any angle).
the process of extraction and prediction will begin by running the script "letter_digit_prediction" and the result will be shown in the form of plot and the predicted text.
there are three other scripts which are responsible for extracting the database form "preprocess_3.py" and training, testing and validating "Digits_neural.ipynb , Letters_neural.ipynb".
nearly 2,000 data for each letter
nearly 1,500 data for each digit
totally about 80,000 handwritten letters and digits from more than 300 different people.
numpy (1.16.4)
tensorflow (1.14.0)
keras (2.2.4)
opencv (3.4)