Tesseract OCR not working anymore for not English ingredients

[monsieurtanuki]

What

• In openfoodfacts-dart we test both GOOGLE_CLOUD_VISION and TESSERACT for ingredients.
• From yesterday, the TESSERACT OCR does not work anymore for non English ingredients, like
  • https://fr.openfoodfacts.org/produit/7300400481588/tartine-croustillante-fibres-wasa in French
  • https://de.openfoodfacts.org/produkt/4260107223344/fritz-kola-bio in German
• In both cases, if I try to OCR the ingredients from the website, it does work.
• In off-dart, it's been working OK for months or even years.

[monsieurtanuki]

Typical API call would be something like:

• POST to /cgi/ingredients.pl
• parameters:

'code': barcode,
'process_image': '1',
'id': 'ingredients_fr',
'ocr_engine': 'tesseract'

[teolemon]

We haven't used Tesseract in production for a while.
Do we map it to Cloud Vision ? Do we reinstall?
@alexgarel @stephanegigandet

[alexgarel]

@teolemon it might be that I did not re-install tesseract on the new server (seems it was not documented). Is it important ?

[teolemon]

no it's not, but we should send back cloud vision results for people who try the tesseract param. I guess ?