B4A Library OCR

This library will extract text from images via OCR.

Requirements:

1- B4A 3.82.

2- Get an API Key HERE (you need to create an account).

3- Internet connection.


How to install:

- Copy the OCR.jar and OCR.xml to your additional libraries directory.

Usage:

Private myOCR As OCR

myOCR.Initialize(Me, "myOCR", "<YOUR_API_KEY>")

myOCR.ExtractFromImage(File.Combine(File.DirAssets, "sample.gif"), "", "scene", 20000)

That's it.

You can upload an image or reference one via a URL, the service will return the extracted text (if any).


NOTES:

- OCR is not perfect, so, expect some weird results or no results at all.

- It is a free service but some quotas apply.


For more details, see the attached project.

 

[Tom Christman]

Really cool

 

[Douglas Farias]

works fine on brasil *-* portuguese linguage perfect *-*

 

[MarcoRome]

NJDude fantastic work. Thank you

 

[Douglas Farias]

@NJDude you know the diference of scene , Photo etc?
I see the coments on your lib, but working i realy dont see diference
what you recomend to use?

 

[NJDude]

The default is "photo", but in some cases you have to use the other ones depending on what kind of picture you are submitting, you will see that in some cases the default (photo) will not return any text and by changing it to scan (or any other) it will, you will have to try.

 

[Douglas Farias]

i go make a chosse option when user go send
and about socket time error?
its possible add on the request a timeout? 60000 seconds etc ?

 

[NJDude]

The default timeout is 20 secs, but I could modify the library to accept a timeout parameter.

 

[Douglas Farias]

thx man, the requests for portuguese is more time *-* sometimes works in 20s , english is very fast *-*
i go wait thx

 

[NJDude]

I don't think that the language has anything to do but I will add the timeout parameter anyway.

 

[JohnD]

Very cool! I fed it a php source code file saved as a gif. It did a great job for a work in progress. Do you know if it works better with a particular Windows font and size? I'd like to play around with it some more.

 

[NJDude]

A new version has been uploaded to the first post, it adds a new parameter: Timeout to the ExtractFromImage method.

 

[NJDude]

Very cool! I fed it a php source code file saved as a gif. It did a great job for a work in progress. Do you know if it works better with a particular Windows font and size? I'd like to play around with it some more.

It is hard to say, OCR can be very picky, but, so far, during testing it did a decent job.

 

[JohnD]

If I can identify any specific preferences or dislikes the software has to fonts and sizes I'll let you know.

 

[NJDude]

One thing I can tell you is that any OCR works better if the image (text especially) is using high contrast, the extraction is better, there's no perfect OCR.

 

[Robert Vermulen]

Hi,

I have tested in android 2.3.6 it is working fine, but in 4.0.4 its showing error " javax.net.ssl.sslpeerunverifiedexception no peer certificate ".

plz help me to fix this.

Thanks

 

[NJDude]

Are you trying on a real device or the emulator?, I don't have that problem on any of my devices.

 

[LucaMs]

I obviously appreciate (and "Like"d) this in your library and your job, NJDude.

But I'm thinking about how many times we rest to external sites for our apps: and if the site in question shut?

I should open a thread for this, although I do not know exactly where.

I'll do it in chit-chat with a reference in the questions

 

[Erel]

You need to use a more recent version of android.jar (under Tools - Configure Paths).

 

[MarcoRome]

Hi NJDude.
One question... but work only with characters latin. Because with russian , chinese characters dont work.
Any idea ?
Thank's