Derin sinir ağları kullanarak el yazısı algılama ve belge analizi

In this work, an application of deep learning techniques to recognize handwritten source code characters is presented. Although there are many works on the handwritten character recognition (HCR) problem, very few have been done about the offline handwritten source code character recognition. The problem includes the recognition of source code specific characters. An application designed and implemented, performing preprocessing, histogram based segmentation and normalization on the scanned documents of exam papers which include codes that were written in C programming language. Constructed dataset includes 7093 source code character samples. This dataset was enriched with character samples from the CROHME database by transforming them to offline samples. With resulting 95 classes of 17748 samples, several models of Convolutional Neural Networks (CNNs) were trained and tested. CNN is a deep learning architecture which is shown to produce state-of-the-art performance rates for handwritten character recognition tasks as in various other computer vision applications. Experimental evaluations gave performance rates between 92.33 percent and 98.82 percent. We conclude that CNN based classifiers are powerful tools for recognition of handwritten source code characters task.