question about performance (inproc, outproc)

  • Last Post 25 October 2018
김태성 posted this 10 October 2018


If I perform inproc and outproc recognition, outproc takes about 50% more time than inproc. Is it possible to check if it is normal behavior? (document text has small and many blocks and letters)

best regard,

Order By: Standard | Newest | Votes
Helen Osetrova posted this 18 October 2018



Could you please describe your usage scenario in more details? Could you also provide us with the sample document?

For faster communication, you can contact directly Please specify in the email your License Serial number and the FineReader Engine build number.


김태성 posted this 20 October 2018

Dear Helen,

In a mixture of Korean and English, the document set is used as input.

For the documents in the input set, we go through the steps of preprocessing -> classification -> recognition. The only difference in the experiment is the difference between inproc and outproc in the process of creating the engine.

Additionally, the image used for input is not supported finereader engine and we pushed into the engine using FREngine::AddImageFileFromStream.

It is presumed that the speed difference is caused by this part. After writing the decoded image to the file system as a temporary file and FRDocument :: AddImageFile is called, but the performance issue isn't solved

It takes about 90 seconds for inproc and 180 seconds for outproc.

Providing images is difficult, but we will think this problem occur general document too.

we use

김태성 posted this 25 October 2018

we fixed this problem using plaintext.

thank you!