Smart OCR for Recognizing Bangla Characters with CRAFT and Deep Learning Models
Md Rakibul Hasan, Anamika Basak Pew, Sanzida Alam, Nafisa Tasnim Rifha, Shamin Yeaser Shams, Farhan Shahriar, Rashedur Mohammad Rahman · 2022 IEEE 13th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON) · 2022
Optical Character Recognition (OCR) is a popular technology for a variety of applications. The purpose of implementing optical character recognition is to turn printed text into editable text. OCR accuracy can be influenced by text preparation and segmentation techniques. Occasionally, it might be difficult to obtain text from an image due to its varying size, style, orientation, and intricate backdrop, among other factors. If the printed or handwritten texts are in Bangla font, the work gets more difficult. Because Bangla language has 50 characters and 24 scrape characters for example, etc. Moreover, in bangla language there are several dipthong or coumpound letter for example, etc. Therefore, even though there are several OCR-related works really good for the English font, for Bangla, we do not have without NLP (Natural Language Processing). So here We have tried to cover both the English and the Bangla font with our proposed model using crafting, VGGNet and various other machine learning algorithms such as LSTM and CTC.