Machine Learning & Model Building (OCR Tool) Work From Home Internship

Optical character recognition (OCR) is the process of converting document images into an editable electronic format. This has many advantages like data compression, enabling search or edit options in the images/text, and creating the database for other applications like machine translation, speech recognition, and enhancing dictionaries and language models. OCR in Indian languages is quite challenging due to the richness in inflections.

Using open-source and commercial OCR systems, we have observed the word error rates (WER) of around 20-50% on printed documents in four different Indic languages. Moreover, developing a highly accurate OCR system with accuracy as high as 90% is not useful unless aided by the mechanism to identify errors. So, we started with the problem of developing 'OpenOCRCorrect', an end-to-end framework for error detection and corrections in Indic-OCR. Our models outperform state-of-the-art results in 'Error Detection in Indic-OCR' for six Indic languages with varied inflections and we have solved the out of vocabulary problem for 'Error Correction in Indic-OCR' in our ICDAR-2017 conference paper. We further improve the results with the help of sub-word embeddings in our ICDAR-2019 conference paper. Demo video for our framework is https://www.youtube.com/watch?v=u9bqUDrGugc

To install the software, you can go to https://github.com/rohitsaluja22/OpenOCRCorrect and follow the instructions given in https://www.youtube.com/watch?v=0hcdlF-zn8E.

There is an immediate demand to keep the softcopy of the Indian preserved texts. Currently, we are targeting Sanskrit. Although the OCR tools available online do a decent job on English texts, they are not optimized for Indic languages. Thus developing an OCR model for the same is our concern. The model should be able to detect text with maximum level accuracy and should be able to draw bounding boxes on each line of the text. Further, in the digitization process of such texts, the second step would be spelling correction and formatting of the text detected by the OCR models.

1. ICDAR 2019 Post-OCR competition: Our team 'CLAM' secured 2nd position in the multilingual PostOCR competition at ICDAR'19. Our model achieved the highest corrections of 44% in Finnish, which is significantly higher than the overall topper (8% in Finnish). Final report: https://drive.google.com/file/d/15mxNO-M9PiXBnffi7MOa8wUw33nj1xBp/view?usp=drive_open) and poster available (https://drive.google.com/file/d/1uuBWu1LQ1QZ49SCgLBoB1er4HpWSzmcx/view.

2. ICDAR2019: You can read the paper here - https://www.cse.iitb.ac.in/~rohitsaluja/PID6011473.pdf

3. ICDAR2017: You can read the paper here - https://ieeexplore.ieee.org/document/8269944

4. ICDAR-OST 2017:
(A) OpenOCRCorrect: you can read the paper here - https://ieeexplore.ieee.org/abstract/document/8270254
(B) Source code for our framework is available here - https://github.com/rohitsaluja22/OpenOCRCorrect.