Georgian Handwritten OCR and Generator

Author: davit soselia
Co-authors: Irakli Koberidze, Sandro jijavadze, Shota Amashukeli, Giorgi Gigauri, Levan Shughliashvili
Annotation:

During the research, a framework for recognizing handwritten Georgian characters using Convolution Neural Networks and transforming the input into another person's (currently Javakhishvili) was developed. for training the CNN model an extensive dataset was created with over 50 000 character samples. for Correction of recognised text 363 201 unique words were gathered. The framework has been deployed in the web and Windows, Linux, iOS operating systems.

Lecture files:

abstract handwritten georgian text recognition and generation [en]
პრეზენტაცია [ka]