Dockerfile.tess4 won't work due to 16.10 no longer being supported by ppa:alex-p/tesseract-ocr

[hernick-qc]

The current version of Dockerfile.tess4 is based on Ubuntu 16.10, but the ppa:alex-p/tesseract-ocr no longer offers tesseract-4 builds for 16.10. However, 17.04 is now supported by the PPA, and simply changing the Dockerfile FROM ubuntu:16.10 to ubuntu:17.04 works great for me.

I'm using OCRmyPDF-tess4 to automatically OCR all documents scanned on our OSA Sharp MFP with PaperCut MF and it works great, the users love it, and the quality of the results are better than with tess3. Only downside, it takes nearly half an hour to OCR a 100 page document on a Xeon E3-1240 V2.

[jbarlow83]

I’ll make the change. There’s not much I can do about the time to OCR. However you may want to ensure that the Docker container runs with access to all CPUs since some configurations only let it see one.

…

On Fri, Oct 6, 2017 at 11:57 hernick-qc ***@***.***> wrote: The current version of Dockerfile.tess4 is based on Ubuntu 16.10, but the ppa:alex-p/tesseract-ocr no longer offers tesseract-4 builds for 16.10. However, 17.04 is now supported by the PPA, and simply changing the Dockerfile FROM ubuntu:16.10 to ubuntu:17.04 works great for me. I'm using OCRmyPDF-tess4 to automatically OCR all documents scanned on our OSA Sharp MFP with PaperCut MF and it works great, the users love it, and the quality of the results are better than with tess3. Only downside, it takes nearly half an hour to OCR a 100 page document on a Xeon E3-1240 V2. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#191>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABvcM1u6of6S-o-YWie6QwyIC4HF8WbDks5spmoOgaJpZM4Pw1E-> .

[jbarlow83]

My fix to this issue is blocked by tesseract-ocr/tesseract#1167

That is, because of this segfault ocrmypdf's test suite will not pass, and so the Docker images will not generated (and wouldn't work anyway).

[jbarlow83]

The issue above also describes the workaround, which is to replace the installed Tesseract 4's tessdata/eng.traineddata with the version from tesseract-ocr/tessdata/eng.traineddata, and for any other language of interest.

[jbarlow83]

Workaround added to v5.4.1