Intel® Processors, Tools, and Utilities
14506 Discussions

tesseract4.0 crash when I use Intel i7-8700K CPU with Ubuntu 18.04



I found tesseract4.0 crashed when I use Intel i7-8700 CPU with Ubuntu 18.04.

tesseract4.0 is a OCR open source. ( GitHub - tesseract-ocr/tesseract: Tesseract Open Source OCR Engine (main repository) )

I used pyocr for wrapping lib for tesseract.

Here is the way I use, PyOCR | python Tutorial

When I call below function several times, it is OK.

txt = tool.image_to_string('test.png'),





But, when I call it over one hundred times, crash happens and PC is reboot.

By the way,

when I use i5-8400 CPU, sometimes error message is shown, but no crash happens.

And, one more thing,

Tesseract uses AVX and SSE, and when I disable them and re-test it, everything is fine. Never crash happens. No error message is shown.

But, you know, running speed is not good.

I think i7-8700 CPU has something with AVX and SSE function which is different from those of i5-8400.

Is there anybody who has a similar experience to me?

Is there any solution I can use tesseract with i7-8700K CPU turning on AVX and SSE.

2 Replies

Hello DrPRO,



Thank you for joining the Intel Community Support.



I understand that Tesseract 4.0 crashes in a system built with an Intel® Core™ i7-8700K Processor in Ubuntu 18.04. I would like to provide further recommendations.



To address this issue correctly, we recommend you to visit these websites as you may need additional documentation to fix this issue.



You can also contact the developers of this open source engine.



I hope this helps.



Wanner G.
0 Kudos

I am experiencing the same problem; running Tesseract on certain files causes my entire computer (with an Intel Core i7-8700 @ 3.20 GHz) to do a hard reset.  I am running openSUSE Tumbleweed with Linux 6.1.1.  In the Tesseract issue tracker, about six other people have all reported the same issue with various Intel CPUs and GNU/Linux distributions.  It's clear that the problem is not specific to Ubuntu or any one GNU/Linux distribution.  It's also pretty clear that this isn't a bug in Tesseract, since a user-level application should not be able to force a hard reset of the computer (without even going through the OS's shutdown sequence).

I suppose the only possibilities left are a Linux kernel bug or a fault in certain Intel CPUs (or the associated microcode), possibly related to AVX2 and/or power management (since disabling AVX2-specific code in Tesseract works around the issue).  So far I have no evidence that it's a kernel bug; no kernel error messages get output to the console or system log before the hard reset.  Can we escalate this problem to an Intel support engineer?  I can provide step-by-step instructions for triggering the problem.

0 Kudos