Hey guys,
we have build an application to automatically extract data from different types of documents. We have a document definition for each type and built a classification tree.
In the application we create a processor object, load the definitions, the classifier and the documents itselfs, then call RecognizeNextDocument and after that we export data to file. So far so good.
As mentioned on page 129 of the user´s guide we would like to speed up the whole process by using only the classification tree. With calling ClassifyNextPage after loading classification tree and documents we get the correct class / definition name. In user´s guide the next step is described as follows:
"Process the information about image category in the way you need."
We tried those steps after classification:
- load the specified definition file
- call RecognizeNextDocument
As a result we got a positive match but unexpected data.
We found out that in this case the processor only recognizes pages 2 - N of that document. Obviously because of the ClassifyNextPage - call before.
If we first reset the processor and then call RecognizeNextDocument our data again is correct.
But: where´s the speedup? According to log outputs it takes twice as much time with that method.
Can someone point to our failure? Or is the described Method - creating processor, loading definitons, classification tree and documents, recognizing - already the fastest way to do that?
regards
Michael
we have build an application to automatically extract data from different types of documents. We have a document definition for each type and built a classification tree.
In the application we create a processor object, load the definitions, the classifier and the documents itselfs, then call RecognizeNextDocument and after that we export data to file. So far so good.
As mentioned on page 129 of the user´s guide we would like to speed up the whole process by using only the classification tree. With calling ClassifyNextPage after loading classification tree and documents we get the correct class / definition name. In user´s guide the next step is described as follows:
"Process the information about image category in the way you need."
We tried those steps after classification:
- load the specified definition file
- call RecognizeNextDocument
As a result we got a positive match but unexpected data.
We found out that in this case the processor only recognizes pages 2 - N of that document. Obviously because of the ClassifyNextPage - call before.
If we first reset the processor and then call RecognizeNextDocument our data again is correct.
But: where´s the speedup? According to log outputs it takes twice as much time with that method.
Can someone point to our failure? Or is the described Method - creating processor, loading definitons, classification tree and documents, recognizing - already the fastest way to do that?
regards
Michael
Comments
1 comment
It seems that your question is related to SDK product, FlexiCapture Engine, which is unfortunately out of the scope of this forum.
To get assistance with FlexiCapture Engine please contact technical support team in your region, see FlexiCapture Help or http://www.abbyy.com/partnerslocator/ for contact details.
Please sign in to leave a comment.