search

jonathanpalma / react-native-tesseract-ocr

Tesseract OCR wrapper for React Native
MIT License
565 stars 172 forks source link

output is not correct in android #35

Closed ghost closed 3 years ago

jonathanpalma commented 6 years ago

Could you please provide more information about this?

Rananjaya commented 6 years ago

I have same issue.

imran1992 commented 6 years ago

Yes.Same here too.... I want to recognize LANG_ENGLISH. I just copy one file eng.traineddata to MyRNApp\android\app\src\main\assets\tessdata. Or Do I need to copy all the files in v3.04 Trained data files

chetankotkar commented 5 years ago

same issue am facing .. I am getting output like below..

2019-02-19 15:36:07.843 21999-22040/com.contentreader E/RNTesseractOcr: g —z é , g a ’ g. a." u» z a ‘g pan u m a in “an an m"“‘\ Oabwiz S zézwai ‘Q °~~ a susitnz“ fl §E£ia

shiprasharma12 commented 5 years ago

Hello Guys,

In Android build using react native, the OCR is not returning the expected result.

Please provide your inputs so that I can move further in this module as I'm stuck with this issue.

Thanks, Shipra Sharma

nguyenvanan18031993 commented 4 years ago

hello guys, Same issue. I get incorrect output when scan Number in captured image from camera. Thanks.

Rananjaya commented 4 years ago

same issue here. output not correct

lutakyn commented 3 years ago

i have the same issue, the text in my image is "I am Three" , the output is k. Is the trained data wrong?

arunahuja94 commented 3 years ago

same issue here. output not correct Any solutions?

wadewadewadewadewadewade commented 3 years ago

I used the 4.0 training data from here: https://github.com/tesseract-ocr/tessdata ; and the result isn't great.

When I use it on an image that has a cartoon zebra in the background partially and some sans-serif green text reading 'ZEBRA' that's huge in the image, with letters spread far apart, I get:

E \ '.
Z EEB'BA

Which isn't great, but is at least better then a handwritten pencil on white paper with the word 'salt', which end up just a slurry of nonsense. My ultimate goal is to recognize handwriting.

Questions:

  1. Would it help if I processed the image after it was chosen to make it just black and white, and smooth out the pixels?
  2. Or maybe there are OCR threshold settings I can tinker with somewhere in the android folder?

Thanks for the help!

lloytynoja commented 3 years ago

@wadewadewadewadewadewade

As project uses tess-two, I think you should use 3.x training data. At least on my app, results are now better (not perfect yet, but haven't explored with image sizes nor other parameters yet). The data is available here (linked in tess-two documentation): https://github.com/tesseract-ocr/tessdata/tree/3.04.00

isfandiarto commented 3 years ago

i have the same problem, for me it seems to work only on image with white font and black background, other than that the result is gibberish. Any advice to this problem

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.