sikuli-driver team mailing list archive

Thread
Date

Re: [Question #273729]: Training tesseract OCR for making Region.click("String") better

To: sikuli-driver@xxxxxxxxxxxxxxxxxxx
From: umesh <question273729@xxxxxxxxxxxxxxxxxxxxx>
Date: Wed, 04 Nov 2015 15:57:45 -0000
Reply-to: question273729@xxxxxxxxxxxxxxxxxxxxx
Sender: bounces@xxxxxxxxxxxxx

Question #273729 on Sikuli changed:
https://answers.launchpad.net/sikuli/+question/273729

Description changed to:
I have my types of font which used in application. Those are not standard font , but modified one. When I use .text() and Region.click ("String") command then sometimes it fail to recognize the string and hence I decided to train the OCR for my types of fonts. I have trained it manually first and replaced trainedData from C:\Users\Admin\AppData\Roaming\Sikulix\SikulixTesseract\tessdata by trainedData that I have created, but sikuli crashed when I execute the command .text() or Region.click("String") command.
Then I trained OCR by using  jTessBoxEditor-1.4 tool and and kept it in previous said location. Sikuli worked this time but results are worse than what had seen by original trained data.
I want to know how you have trained OCR, by using what images/ fonts and dictionary files ? This will help me to train OCR to detect my types of fonts.

Thanks in advance.

-- 
You received this question notification because your team Sikuli Drivers
is an answer contact for Sikuli.