adaptech-cz/Tesseract4Android

Not working with some traineddata files for tesseract 4

nkhdiscovery opened this issue · 1 comments

Hi guys, great job done! :D

I used your library for a while and it is working well, until recently that I tried to use with this traineddata file:
https://github.com/Shreeshrii/tessdata_shreetest/blob/master/fas-minus-float.traineddata

I extracted the mentioned traineeddata file and the .version file says:
4.0.0-beta.1-232-g45a6:fas:minus20180518:from:4.00.00alpha:Arabic:synth20170629

while for the eng.traineddata shipped with this repo, the .version file says:
Pre-4.0.0

Is my fas-minus-float.traineddata version is right? Can it be used with your library?

I soon provide the error thrown on my android device, sorry that I cannot provide it at the moment. Thought maybe the version of my traineddata is not compatible at all so that error is not important.

Hi, I think problem is not in "version" but in the fact that your tessdata uses float model (as they say in repo, and as it is in filename), but models that I link from this repo are integer models. Probably Android requires use of integer models only.

Maybe you will be able to convert that float model into integer model with some combine_tessdata script, but I can't help you with that.