Assertion failure when using the legacy ara.traineddata in latest tesseract version.

[girikum]

I downloaded the legacy arabic model file from your documentation here: https://tesseract-ocr.github.io/tessdoc/tess3/Data-Files.html

When I use the model in the latest tesseract version with OEM mode as 0, I see an assertion failure.

tesseract inputImage.jpg outputText -l ara --oem 0
mgr->GetComponent(TESSDATA_INTTEMP, &fp):Error:Assert failed:in file src/classify/adaptmatch.cpp, line 539
Illegal instruction

Is this a bug in the tessdata file or the latest tesseract version?

[stweil]

It's a bug that the latest Tesseract raises an assertion instead of reporting an old format which is obviously no longer supported. Why do you want to use a legacy model for Arabic? I don't expect that it will work good.

[amitdo]

Duplicate of tesseract-ocr/tesseract#3929.