Offline Pashto OCR Using Machine Learning
Sultan Ullah, Tehmina Enayat, Noor Nadeem, Ikram Ud Din, Yousaf Saeed, Muhammad Junaid · 2019
For a long time, many techniques have been used for Optical Character Recognition in image processing. But Pashto optical character recognition is less developed area. Because of Pashto cursive nature, writing direction from right to left, and its character change its shape when it is placed at the start, middle and end makes the development of Pashto OCR a complex task. The proposed system takes an image with Pashto text input. After that process it and pass through its modules and generate an editable Pashto text file. More than 5000 images in dataset is use to classify the characters using Sequential Minimal Optimization classifier. An accuracy of our proposed system is approximately 92 Percent.