Georgian handwritten letters recognition

Author: Grigoli Babajanian
Keywords: georgian ocr, neural network, machine learning
Annotation:

The aim of the master thesis is to discuss how the Optical Character Recognition (OCR) technology works, what technical problem it can solve, how it works and what modern approaches are available. Specifically, the paper discusses the essence of the problem of recognizing Georgian handwritten letters and one of the solutions. The goal was to achieve a stable and effective implementation step, which could then be easily and reliably adapted to different test environments to enable automatic model testing and implementation. While working on the technology of recognizing Georgian optical symbols, we aimed to make it easy for our embedded model to perceive different variants of Georgian handwritten letters. Medium resolution should be sufficient to identify the letter on the image. The authentication system should work quickly and steadily.