Build from source failed to recognize arabic

[Essam Zaky]

Hi dears

i had build tesseract and training tools from source for windows and VS2015

when recognize English page it succeeded

but when try to recognize arabic page it fails

C:\Users\emz\tesseract\build\bin\Debug>tesseract eurotext.tif eurotext -l eng

Tesseract Open Source OCR Engine v4.00.00dev with Leptonica

Page 1

C:\Users\emz\tesseract\build\bin\Debug>tesseract sample1.tif sample1 -l ara

Error: LSTM requested, but not present!! Loading tesseract.

tessdata_manager.SeekToStart(TESSDATA_INTTEMP):Error:Assert failed:in file C:\Us

ers\emz\tesseract\classify\adaptmatch.cpp, line 537

what could be the reason of the error

Best regards

Essam

[universal reseller]

​send output of
tesseract --list-langs

in cli

[Essam Zaky]

Hi @.peiman

thanks for reply

i found the problem

I was installed old build for v4 from DanBolomBerg site

and the TESSDATA_PREFIX was refering to old version with cube

now i updated the TESSDATA_PREFIXin system enviornment to the new downloaded data it's working

Thanks again

[universal reseller]

​what is accuracy of result for you!?

[Essam Zaky]

For the sample images i used

The accuracy for english is good

but for arabic the cube is still better than current LSTM