Wednesday, August 20, 2008

Status of Bangla character segmentation

This post is specially for those who are concern about our current status of Bangla character segmentation. I would like to show one example:


Figure 1: Input image


Figure 1: Segmented color image

Yet there is some modifications need to do.

5 comments:

জয়ন্ত said...

গত দুই মাসে পোস্ট থেকে বুঝতে পারছি, কিছু ডেভল্পমেন্ট হয়েছে, কিন্তু বোধ হয় পুরোটাই লিনাক্সের জন্যই, উইন্ডসের জন্য কোনো ভার্সন করা যায় না? আমি বর্তমানে এটা ব্যবহার করছি, কিন্তু train করার পরও ঠিক ঠাক output আসে না।আমি আলফা ভার্সনটা ব্যাবহার করছি

Md. Abul Hasnat said...

Dear Joyonto da,
many many thanks for your comment. If you can send a scan picture of your document then it will be easier for me to comment. The output is still not satisfiable (around 60 - 70 %) for good quality documents. I am continuing my work and I am hopeful. Please pray for me. I may need your help in future about some issues.

জয়ন্ত said...
This comment has been removed by the author.
জয়ন্ত said...

Thanking you for respond and great utility to you and all of your CRBLP team mate. I am from Calcutta. I am very proud to be a bengalee and I love to write Bengali anything in web. I am now busy in Bengali Wikipedia and wiki source. So I need a good Bengali OCR, Spell Checker, and Bengali Machine Translator. I know there are lots of work remain regarding Bengali Computing and you are the pioneer of Bengali computing. Your last OCR release on 2006 and I am waiting for next good version. I know that you do OCR development work as volunteer and it will take time. My document is not so good, still I am sending to you. Thanking you.

Md. Abul Hasnat said...

Dear Joyonto da,
thanks for your appreciation. At present my main focus is to develop a standard OCR. I am completely focusing on OCROPUS project to integrate Bangla Language recognition support. After completing the basic tasks with a specific font I will move to work with all other favorite fonts. So, in that case I need your support. It is really encouraging for me that you are always concern about the OCR to work. Thanks again.