UWPOCR - Windows Platform Optical character recognition API Implementation

[drrehak]

i am getting an unterminated string error. This is from Console:

C:\Users\drreh\Desktop\autoit\UWP_test.au3 (152) : ==> Unterminated string.: 
<meta name="optimizely-datafile" content="{&quot;version&quot;: &quot;4&quot;, &quot;rollouts&quot;: [], &quot;typedAudiences&quot;: [],
........
&quot;anonymizeIP&quot;: true, &quot;projectId&quot;: &quot;16737760170&quot;, &quot;variables&quot;: [], &quot;featureFlags&quot;: [], &quot;experiments&quot;: [{&quot;status&quot;: &quot;Running&quot;, &quot;audienceIds&quot;: [], &quot;variations&quot;: [{&quot;variables&quot;: [], uot;, &quot;key&quot;: &qu

I added the "........." in middle to abbreviate

[Danyfirex]

@drrehak are you using Windows 10?

 

Saludos

[drrehak]

15 minutes ago, Danyfirex said:

@drrehak are you using Windows 10?

 

Saludos

I am an idiot. I think I miscopied the UDF. lol

[drrehak]

omg. I have tried scanning this image with Tesseract with no luck. Now I am using UWPOCR. the 44 does NOT want to be recognized.

The sample image is attached. I have it open in Windows Snipping tool. any ideas?

#include <GUIConstantsEx.au3>
#include <ScreenCapture.au3>
#include <UWPOCR.au3>
WinActivate ("Snipping")
sleep(400)
    _GDIPlus_Startup()
    Local $hHBmp = _ScreenCapture_Capture("", 0, 0, 500, 500)
    Local $hBitmap = _GDIPlus_BitmapCreateFromHBITMAP($hHBmp)
    _WinAPI_DeleteObject($hHBmp)
    Local $sOCRTextResult = _UWPOCR_GetText($hBitmap, Default, True)
ConsoleWrite ($sOCRTextResult)
     Local $hGUI = GUICreate("GDI+ test", 800, 800, -1, -1)
    GUISetState()

    Local $hGraphics = _GDIPlus_GraphicsCreateFromHWND($hGUI)
    _GDIPlus_GraphicsDrawImage($hGraphics, $hBitmap, 0, 0)

    While 1
        Switch GUIGetMsg()
            Case $GUI_EVENT_CLOSE
                ExitLoop
        EndSwitch
    WEnd

    ;cleanup resources
    _GDIPlus_GraphicsDispose($hGraphics)
    _GDIPlus_BitmapDispose($hBitmap)
    _GDIPlus_Shutdown()
    GUIDelete($hGUI)

 

[ptrex]

Danyfirex,

Great job as usual ! 👌

Now we can trash the old fashioned COM OCR library

https://support.microsoft.com/en-us/topic/install-modi-for-use-with-microsoft-office-2010-4fbd3076-6d01-9cb7-c574-3bbabc9eead9

Which I still use daily as an integrated component of of Greenshot...

https://audministrator.wordpress.com/2017/08/07/greenshot-adding-ocr/ 

Thanks for sharing !!

[Danyfirex]

Updated. Added new function to get the Bounding Rect of the words.

 

Saludos

[funkey]

Hi,

 

strange, but this looks a bit wrong for me.

 

[mLipok]

1 hour ago, funkey said:

 

@funkey is this your own demo ? Or is this part of  @Danyfirex UDF ?
 

Edited February 7, 2022 by mLipok

[funkey]

I have nothing changed. I just used the files from GitHub.

Maybe it is because german language?

[mLipok]

Oh I see . I was not aware that  Examples/04 - Get OCR Words To 2DArray.au3 was added.

[mLipok]

Check system settings: zoom and smoothing.

Do they affect the action?

[Danyfirex]

@funkey I don't know what could be the issue. try with some German text and check if you get correct bounding box.

 

Saludos

[funkey]

Hello, I tried a bit and then I changed the font and it works as expected. So it sems there is a problem with 'Comic Sans MS' on my PC. Most other 'easy readable' fonts work well with the bounding rects.

Thanks for your great UDF!!

[Danyfirex]

@funkey I'll update the font. Thank you 😊

[funkey]

Hi @Danyfirex, I tried a bit with your fantastic UDF and wanted to use the angle. But it seems not to work.

I only get values like this: '1.37102911403396e-311'

 

I have no clue, what is going wrong, but I saw this can be a nullable double value. I don't know how to handle this. Maybe this is not a problem as yo declare $iAngle to 0, and if TextAngle is called with null in angle then it will not be assigned.

 

BTW: To keep the angle in @extended your wrapper functions should look like this:

Func _UWPOCR_GetText($sImageFilePathOrhBitmap, $sLanguageTagToUse = Default, $bUseOcrLine = False)
    Local $oErrorHandler = ObjEvent("AutoIt.Error", __UWPOCR_ErrorHandler)
    #forceref $oErrorHandler
    _UWPOCR_Log("_UWPOCR_GetText")
    Local $sRes = __UWPOCR_GetText($sImageFilePathOrhBitmap, $sLanguageTagToUse, $bUseOcrLine)
    Return SetError(@error, @extended, $sRes)
EndFunc   ;==>_UWPOCR_GetText

Func _UWPOCR_GetWordsRectTo2DArray($sImageFilePathOrhBitmap, $sLanguageTagToUse = Default)
    Local $oErrorHandler = ObjEvent("AutoIt.Error", "__UWPOCR_ErrorHandler")
    _UWPOCR_Log("_UWPOCR_GetWordsRectTo2DArray")
    #forceref $oErrorHandler
    Local $aRes = __UWPOCR_GetText($sImageFilePathOrhBitmap, $sLanguageTagToUse, False, True)
    Return SetError(@error, @extended, $aRes)
EndFunc   ;==>_UWPOCR_GetWordsRectTo2DArray

 

Edit: Forgot to say that @extended only can hold Int32 values, so you loose precision, when using it for the angle.

Edited February 11, 2022 by funkey

[jimg]

I'm trying to use UWPOCR and I'm getting some odd results:

1) The phrase "XML" or "xml" is being ignored by the OCR as is "PDF".  It's almost like "hide known file extensions" is somehow happening.  I have it off in WindowsExplorer but maybe there's a more global setting?

2) A double underline "__" is causing a line break or something.  If I OCR a list of phrases containing that string, the returned text is jumbled.

I've attached a sample jpg and the resulting text.  I've put an Explorer window just below my target data so you an see the text is otherwise list.txtrecognized.

$imagefilepath1="./scrape.jpg"
$list = _UWPOCR_GetText($imagefilepath1)
 

I've toyed with all sorts of area values to eliminate the vertical line artifact, change size, etc. with nothing working.  Elsewhere I suffer from the "__" issue but I might be able to work around that if I can get "XML" to translate.

I'm hoping there's some parameter buried in the script that allows some configuration options?

[jimg]

I've tried adjusting brightness/contrast with no improvement.

I created a Notepad image that's roughly the same and it translates fine, so there's something more diabilical.  The file extentions issue might be a red herring.

[Danyfirex]

It may need some image processing I think. Some threshold,medianBlur,dilate etc.

 

Saludos

[jimg]

I did a Gimp2 "threshold" and messed with the brightness/contrast with no improvement.  I'll try some more.

Tried scaling as well.  The string "Autoit" is a lot thinner and dimmer and it is seen.

Edited March 27, 2022 by jimg

[SaeidN]

Farsi  (Right to Left) language tag is "fa" and I give it a png file with Farsi text with normal font. It returns nothing. But English works perfect. Any idea to fix it?