Machine Learning Empowered Urdu Characters Recognition Mechanism

The study was conducted on digitally written Urdu characters. After a little preprocessing (resizing the images and separating different characters based on the classes), we applied different features extraction techniques and artificial intelligence algorithms on the same dataset and compared the results, to successfully predict digitally written Urdu characters. We managed to classify the data with an accuracy of 98%. But of course, these are results from the best combination of algorithms, feature extraction, and hyperparameter combination. There is still a huge issue of recognizing handwritten characters. Much work has been in previous years to successfully recognize characters but most of the research was done in English characters or digits and very less work is done in recognizing Urdu characters. One of the main reason for this is the same characters has different shapes depending on their position in the word. Moreover, some properties of Urdu like calligraphic nature also cause a lot of problems. To cover all these problems, we studied several feature extraction techniques and algorithms to drive the best possible results.