OCR speed in V8 slow

Discussion for the End User use of OCR in PDF-XChange Editor and Viewer

Moderators: TrackerSupp-Daniel, Tracker Support, Paul - Tracker Supp, Vasyl-Tracker Dev Team, Chris - Tracker Supp, Sean - Tracker, Ivan - Tracker Software, Tracker Supp-Stefan

Post Reply
User4455
User
Posts: 54
Joined: Tue Nov 06, 2018 7:59 am

OCR speed in V8 slow

Post by User4455 »

Dear Support-Team,

after having updated from V7 to V8, OCR appears slower in the Editor and in the Tools than before the update (instead of faster, as advertised). Unfortunately, I cannot provide an actual time difference, because I do not have V7 installed anymore.

I have the "Enhanced OCR" plugin installed.

Please let me know, if you have an idea on that issue.

Note, that I need the result to be a "searchable image" (in my german version "Durchsuchbares Bild"), as I need to have the original information accessible at all times.

Cheers and Thanks in advance!
User avatar
TrackerSupp-Daniel
Site Admin
Posts: 8436
Joined: Wed Jan 03, 2018 6:52 pm

Re: OCR speed in V8 slow

Post by TrackerSupp-Daniel »

Hello usser4455,

Might I ask if you are finding this with the same documents or different ones? when running a test with the exact same original document, (15 pages of only scanned content) in V8, I completed a an OCR with the below settings, in 1:25,
image.png
Conversely, In V7, with the below settings, as similar as I could get between the two versions, it took 6:08 to complete OCR on the same document:
image.png
If you are finding that you believe the OCR is taking longer, you can download the Portable version here, and do a comparison test with the same document. Note that you will need to run one test at a time, as the portable and the installed version will not always want to run side by side.

Kind regards
Dan McIntyre - Support Technician
Tracker Software Products (Canada) LTD

+++++++++++++++++++++++++++++++++++
Our Web site domain and email address has changed as of 26/10/2023.
https://www.pdf-xchange.com
Support@pdf-xchange.com
User4455
User
Posts: 54
Joined: Tue Nov 06, 2018 7:59 am

Re: OCR speed in V8 slow

Post by User4455 »

Dear Daniel,
thank you very much for your response.

In fact, my feeling of different speeds was very much subjective. I have now given it a try and my results are as follows (pdf-document is attached):

Portable Editor V8: 25s

Installed Editor V8: 39s

Installed Tools V8: 42s

all had the following settings (Enhanced OCR was enabled; I did run one test at a time):
Unbenannt.png

I also had the chance to give it a try on an installed editor V6, but on a different machine. I set it to "English, German" and "Medium", too. It took

12s !!!!

I assume, you do not provide legacy versions of the portable version (or of any product) so that I could perform further tests?

Any further thoughts? Thank you!
Attachments
EP1243825A1.pdf
(391.52 KiB) Downloaded 138 times
User avatar
TrackerSupp-Daniel
Site Admin
Posts: 8436
Joined: Wed Jan 03, 2018 6:52 pm

Re: OCR speed in V8 slow

Post by TrackerSupp-Daniel »

Hello User4455,

Thank you for the details, I will run some comparisons over here myself and report this to the Dev team as soon as I can reproduce it.

Regarding previous versions of the software, we only offer portable versions for the Editor, none of our other softwares can be run portably. But you can find the previous versions of any of our products by clicking on the "previous builds" link beside it on our downloads page:
https://www.pdf-xchange.com/product/downloads
image.png
These go back many versions, including V6. We have no requirements for our clients to use the latest version, and we support users who wish to run in legacy versions if they so choose, unlike some competitors. The only limitation is that we cannot resolve existing bugs in those old versions.

kind regards,
Dan McIntyre - Support Technician
Tracker Software Products (Canada) LTD

+++++++++++++++++++++++++++++++++++
Our Web site domain and email address has changed as of 26/10/2023.
https://www.pdf-xchange.com
Support@pdf-xchange.com
User4455
User
Posts: 54
Joined: Tue Nov 06, 2018 7:59 am

Re: OCR speed in V8 slow

Post by User4455 »

Dear Daniel,

thank you for the hint on the legacy versions. This is indeed useful and a policy I very much appreciate!

I have now tried the Editor V7 portable (latest built) with the same document. It took 44 seconds. Out of curiosity, I ran the others again:
portable Editor V8: 43s
installed Editor V8: 42s
Tools: 43s

Well, it is difficult to put my finger on it... In fact, the performance appears not reproduceable. Note in my last post "Portable Editor V8: 25s", now 43s! (though I cannot guarantee that I havent made a mistake back then).

However, any time differences detected on my end are far smaller than the one reffered to in your post of May 17, 2019 6:20 pm. The times measured today are essentially the same, which would disprove both my subjective feeling and the advertized increase in speed. :( (Note that this will not stop me from loving your products!)

I hope my information could help you in some way. I can perform further tests if you ask me to.

Cheers
User avatar
TrackerSupp-Daniel
Site Admin
Posts: 8436
Joined: Wed Jan 03, 2018 6:52 pm

Re: OCR speed in V8 slow

Post by TrackerSupp-Daniel »

Hello User4455,

This similarity likely caused by the fact that the new OCR is actually doing more work than the old OCR was. You have selected to detect and fix deskew which adds some extra steps into the process which the old OCR would not do automatically. To get an overall range, I ran a few more tests and recoded the numbers, Each test was performed with the document you attached to your earlier post to offer some consistency and averages.

In all tests the times was started in the same instant that I clicked OK in the respective OCR dialog, and ended as soon as the processing dialog disappeared and you are able to return to editing the document. All test times are in seconds, and were performed on the portable (32bit only) version of the Editor with both of the default English and German/Deustch language packs selected. Note that results vary based on a number of factors both in the documents, and your PC hardware.

V6 avg - 41.5 seconds
-Perserve Original content and add text layer (no deskew) - avg - 38.88
40.32
36.36
32.97
-Create new searchable PDF, (300 quiality, deskew on) - avg - 44.03
45.84
44.32
41.93

V7 avg - 40.6 seconds
-Perserve Original content and add text layer (no deskew) - avg - 35.44
35.62
34.45
36.27
-Create new searchable PDF, (300 quiality, deskew on) - avg - 45.66
48.43
46.63
41.92

V8 avg - 20.8 seconds
-Searchable image (no deskew detection/fixing) - avg - 21.60
24.70
21.05
19.05
-Searchable image (yes deskew detection/fix) [identical to your PDF-Tools settings above] - not included in average
25.09
19.45
19.76
-Searchable image, (yes deskew detection/fix, and create a new document) - avg - 19.99
20.34
19.33
20.30

Overall, the V8 should fairly consistently be almost twice as fast as both V6 and V7 were, but choosing additional options can slow down the process, as can other outside factors, for example, running OCR in both PDF-Tools, and the Editor simultaneously will likely slow down both instances considerably.

Kind regards,
Dan McIntyre - Support Technician
Tracker Software Products (Canada) LTD

+++++++++++++++++++++++++++++++++++
Our Web site domain and email address has changed as of 26/10/2023.
https://www.pdf-xchange.com
Support@pdf-xchange.com
User4455
User
Posts: 54
Joined: Tue Nov 06, 2018 7:59 am

Re: OCR speed in V8 slow

Post by User4455 »

Hi Daniel,
thank you for your effort and excuse my late response.

I note that according to your measurements V8 is roughly twice as fast as V7. I am sorry, but I cannot reproduce this.

I have conducted a final test:

V7 portable with the following options: 44 sec.
V7.png

V8 installed with the following options: 38 sec.
V8.png

One more with V8 installed and detect and fix deskew enabled: 38 sec.


Well, at least now I know that V8 is not slower than V7 (as I initially feared) but is in fact a little faster on my system. I think I will keep using the deskew-feature since it is supposed to improve accuracy of recognition.

Thanks again!
User avatar
TrackerSupp-Daniel
Site Admin
Posts: 8436
Joined: Wed Jan 03, 2018 6:52 pm

Re: OCR speed in V8 slow

Post by TrackerSupp-Daniel »

Hello User4455,

I am glad that the speed is at least a marginal improvement, I am a bit bewildered as to why there is such a large different between our machines when performing the same actions on the same file, but I suppose that is simply a difference in how the engines are handled on differing hardware.

Nonetheless, I hope you enjoy the new version!
Have a great day.
Dan McIntyre - Support Technician
Tracker Software Products (Canada) LTD

+++++++++++++++++++++++++++++++++++
Our Web site domain and email address has changed as of 26/10/2023.
https://www.pdf-xchange.com
Support@pdf-xchange.com
Post Reply