Bengali Digit Recognition from Real-World Images

  • File:Student-Ambassador-Debdyut-Poster.pdf
  • Size:323.57 KB


Debdyut Hajra

Extraction of text from images has important potential applications, such as visual aids for the blind, guides for tourists, and more. Such a system is still far from reality for many Indian languages like Bengali. This project deals with the development of a robust image preprocessing pipeline to effectively extract isolated digits from an image, and then translate the digits to the desired format. The project also explores how to build an efficient deep learning model to predict the digits by training on standard publicly available datasets.

For more complete information about compiler optimizations, see our Optimization Notice.