Recognize cursive english letters from Image or Photos

54 views
Skip to first unread message

Rajamani Giridharan

unread,
Feb 26, 2018, 10:33:32 AM2/26/18
to tesseract-ocr
Hi,
Tesseract Version 3.04.01 is unable to read cursive text from the product labels (image/photos) like Kelloggs, Nivea, etc.

The tesseract code downloaded from github is returning 'null' results. I tried changing the size of the image and also increased the resolution,but none of them is returning the result. 
Also I tried pre-processing using OpenCV libraries, but the results are none.

Kindly help. I have attached a sample input file for the text extraction.

Thanks and Regards,

Giridharan R
kellogs special.jpg

ada...@turningcloud.com

unread,
Feb 27, 2018, 12:35:14 AM2/27/18
to tesseract-ocr
Hi Rajamani

The issue you might be facing is that Tesseract Reads text for the Fonts it is trained for. The K in kellogs is not in a proper format which can't be read by Tesseract.

Hope this helps.
Regards
Adarsh SHUKLA

Rajamani Giridharan

unread,
Feb 27, 2018, 4:21:35 AM2/27/18
to tesseract-ocr
Hi Adarsh,

Thanks for your response.

I understand the same. I would like to request the Group for some hint or sample input files to train Tesseract ML engine for recognizing cursive letters and other different kind of fonts.

I have already referred to 'Training Tesseract' website. But I am unable to proceed without some sample input training files for 'texts and fonts' that come up in photo images.

Taking this opportunity to request all in the group to help me on the same. 

Thanks,
Giri
Reply all
Reply to author
Forward
0 new messages