Project Goal

Recognizing Thai handwritten text has been a very challenging problem in the field of computer vision (even Google hasn't been able to perfect this). Achieving success in recognizing Thai handwritten text will lead to a big step in computer vision studies. Moreover, such success will help developing computer vision applications which relate to real-life activities such as document transcribing automation, Thai handwritten text translation, and more.

Project Scope

This project aims to provide a Python library for recognizing 1-line Thai sentences from an image into a string of characters using image processing for pre-processing and Convolutional Neural Network for recognizing characters.

Tasks

1. Develop a Convolutional Neural Network (CNN) model for recognizing a single Thai character from an image.

2. Deliver characters in the input image to the CNN model and process the results so that the correct (or at least near-correct) string can be determined.

Project Status

1. Chawan Piansaddhayanon (CP-Chulalongkorn University)

2. Nattaphon Thammarcheep (CP-Chulalongkorn University)

3. Phon Chitnelawong (CP-Chulalongkorn University)

Report (in Thai)