Speed vs Accuracy and API to check a file is OCRed

  • 1.3K Views
  • Last Post 04 June 2013
Workpower posted this 31 May 2013

Hi Users\Support,

I have few queries: 1. Does there exist an API to identify if a file is OCRed or does the engine just runs through any file provided and OCRs it? 2. Is there any API specific available to process an image for Accuracy rather than speed and vice a versa?

Thanks.

Order By: Standard | Newest | Votes
Raj posted this 31 May 2013

Workpower 1. You make the necessary APIL calls to OCR an image you provide as input and indicate how you want to save the results. If you provide the engine a file that is not an image or it cannot open it raises an exception. 2. There are hundreds on configurations that you can set through profile files. It is based on these configurations that you tell the engine how far it should go as regards to accuracy or speed.

Workpower posted this 03 June 2013

Hi Raj,

Thanks for the response. Pardon me if the question does not make sense or is too novice I understand the unsupported format but what I really was looking for if ABBYY has something ( checkIfSearchable etc / non cloud).

In ref to profile files options could you please point me to any API setting or reference witrh regard to same.( I am purely using cloud sdk with API ref as on the portal)

Thanks.

Raj posted this 04 June 2013

Hi @Workpower

I've not used the Cloud SDK but the FineReader 10 SDK. As for profile files they are same as the plain old INI files in windows, for example

private static FREngine.IEngine engine = null;

customprofile = @"c:\somelocation\MyCustomProfile.ini"
engine.LoadProfile(customprofile);

The contents of the MyCustomProfile.ini can be something like below if you want to export your recognise text as UTF-8.

[TextExportParams]
EncodingType = TET_UTF8

So when you export a document as below the engine automatically takes the profiles you' set.

document.Export(@"c:\someoutput\myfile.tzt", FREngine.FileExportFormatEnum.FEF_Text, null);

I haven't consulted the cloud sdk but I expect it to work along the same line as above. You can find samples link text

Anastasia Galimova posted this 04 June 2013

Hi Workpower,

  1. Unfortunately, there is no a method that checks if the input PDF is already searchable. Could you please describe a scenario where you need it?
  2. In Cloud OCR SDK there is no a separate option to specify if accuracy or speed should be prefered, but you can increase processing speed by calibration the parameters of the method you use. For example, if you use processImage method, you can set correctOrientation=false and correctSkew=false (if it suits for your input images).

In FineReader Engine 10 it is possible to set if accuracy or speed should be prefered by using the corresponding predefined profile (the details can be found in FRE10 Help > Index > Working with profiles.

  • Liked by
  • Katia Sirotina
Close