You are viewing limited content. For full access, please sign in.

Discussion

Discussion

Maximum PPM full page OCR in Quickfields with large CPU count.

posted on March 26, 2014

I am trying to determine how expandable the QF product can be.

 

If there is anyone who has tried to maximize their PPM for full page OCR I am curious what numbers you have come up with.

 

I can get up to about 17 PPM full page OCR with a quad core i5-3230M.

 

It would be helpful to know how the product performs across 16+ cores and if anyone has reached a software limitation when throwing more and more cores at it through distributed processing. Any stories or findings would be helpful. Thanks!

 

1 0
replied on May 24, 2014

Hi

 

I also have similar isue and like to optimize the speed of Quick Field session.   I only have one folder to watch.  I was wondering, is it possible to dublicate my existing QF session and run it with QF agent so they both can monitor one folder.    Willl I have any issue? Will it increase the speed?  We have 4 Core processing.

0 0
replied on May 27, 2014

The sessions would end up trying to process the same documents and get errors trying to access images.

0 0
replied on May 27, 2014

Thanks Miruna

0 0
replied on May 27, 2014

You could have a scheduled script to split the documents into 4 folders before the sessions run.

0 0
replied on March 26, 2014 Show version history

OCR only uses one CPU. Multiple cores are only relevant when Quick Fields Agent is used (obviously, a multiple CPU machine will be faster than a single CPU machine since the OS and other applications will not compete for resources with Quick Fields). Depending on the image resolution,how much text there is on the image and disk speed, that might be a reasonable speed.

 

Edit: Distributed Processing improves on that scenario but spinning up multiple concurrent OCR jobs. However, in Quick Fields, processing is done one page at a time.

1 0
replied on March 26, 2014

Very interesting. I was using Quick Fields GUI to see the PPM without knowing it was limiting my core usage. This is assuming a distributed processing / agent scenario and tens of thousands of pages per day.

0 0
replied on March 27, 2014

Quick Fields Agent can run sessions in parallel as well to take advantage of multiple cores. If you only need OCR, then distributed processing is probably the better solution.

0 0
replied on March 27, 2014

If QF Agent is only running one session, does it only have access to one core?

0 0
replied on March 27, 2014

Yes, Quick Fields Agent is just starting off QF to run the session. It does not do anything different than you would do running it manually.

 

If your documents can be segmented off by folder or some other way, you could run multiple copies of the same session in parallel in QFAgent to take advantage of multiple cores.

0 0
replied on March 27, 2014

Ok, we may need to use Workflow / Dist Proc instead in this case. Does Dist Proc  have any of the local enhancement features of QF, such as line removal, despeckle, etc?

0 0
replied on March 27, 2014

Yes, it offers the same image enhancements you see in the OCR dialog in the Client (deskew, despeckle, line removal and rotate).

0 0
replied on March 27, 2014

Thanks, this is good info. It appeared that QF was using all cores from task manager. I will setup a test with Dist Proc and see if there is improvements.

0 0
You are not allowed to follow up in this post.

Sign in to reply to this post.