Can we get more information about suspicious characters?

  • Last Post 18 July 2012
rblasco posted this 21 June 2012


is it possible with Cloud OCR SDK to get the list of possible candidates plus their confidence value for a suspiciuos charParam as it can be done with Abby Fine Reader Engine 10? If not, is there a chance to include this feature in future?


Order By: Standard | Newest | Votes
Vasily Panferov posted this 21 June 2012


Right now it is not possible. Adding this functionality to the API isn't difficult.

What API call are you using? processImage or processTextField/processFields?

  • Liked by
  • rblasco
rblasco posted this 21 June 2012

Hi Vasily,

we call to the processImage action and retrieve the results in XML format. If you can add this feature to the API it would be very usefull for us, can you confirm me it?


Vasily Panferov posted this 22 June 2012

Ok. It will be available in the next release of production system. We'll notify you.

  • Liked by
  • rblasco
rblasco posted this 22 June 2012

great! thanks

Vasily Panferov posted this 02 July 2012

Hi again,

There is an additional parameter to processImage and processDocument. Add xml:writeRecognitionVariants=true to url parameter string. The documentation on site will be updated soon.

rblasco posted this 02 July 2012

ok we will try it, thanks a lot

rblasco posted this 05 July 2012

Hi again,

this new parameter works fine but what about the charConfidence attr for each variant? is not possible to include it?

thanks for your efforts

rblasco posted this 11 July 2012


should I assume that we can not count on it?

rblasco posted this 12 July 2012

Hi Dmitry,

actually when we set the new parameter xml:writeRecognitionVariants to true we get an output like this:

<charParams l="322" t="155" r="330" b="168" suspicious="1"><charRecVariants>

Our question is, can we get a charConfidence attribute for each charRecVariant element as we can when using Abby Fine Reader Engine 10?

<charParams l="964" t="368" r="1008" b="434"> <charRecVariants> <charRecVariant charConfidence="13">0</charRecVariant> <charRecVariant charConfidence="13">O</charRecVariant> <charRecVariant charConfidence="13">o</charRecVariant> <charRecVariant charConfidence="13">Ò</charRecVariant> <charRecVariant charConfidence="13">Ó</charRecVariant> <charRecVariant charConfidence="13">ò</charRecVariant> <charRecVariant charConfidence="13">ó</charRecVariant> <charRecVariant charConfidence="-1">l·l</charRecVariant> </charRecVariants>n</charParams>


Dmitry Me posted this 12 July 2012

This is not implemented at the moment and there's a reason for that - it's rather hard to decide how to interpret the confidence level, specifically how to set a threshold to decide whether a character variant is worth considering or not.

rblasco posted this 12 July 2012

ok, thanks for your response

Dmitry Me posted this 18 July 2012

We've deployed the changes - now when 'xml:writeRecognitionVariants" is specified you get the most detailed XML possible (as in Extended enum value in FineReader Engine).

rblasco posted this 18 July 2012

nice! great job thanks