Daniel Reetz, the founder of the DIY Book Scanner community, has recently started making videos of prototyping and shop tips. If you are tinkering with a book scanner (or any other project) in your home shop, these tips will come in handy. https://www.youtube.com/channel/UCn0gq8 ... g_8K1nfInQ

Scan Tailor "Experimental"

Scan Tailor specific announcements, releases, workflows, tips, etc. NO FEATURE REQUESTS IN THIS FORUM, please.
Tulon
Posts: 687
Joined: 03 Oct 2009, 06:13
Number of books owned: 0
Location: London, UK
Contact:

Re: Scan Tailor "Experimental"

Post by Tulon » 19 Aug 2015, 07:12

Keep in mind simultaneous processing of multiple pages is only available in 64bit version of ST experimental.
Scan Tailor experimental doesn't output 96 DPI images. It's just what your software shows when DPI information is missing. Usually what you get is input DPI times the resolution enhancement factor.

d14b0ll0s
Posts: 31
Joined: 17 Aug 2015, 19:37
Number of books owned: 3000
Country: Poland

Re: Scan Tailor "Experimental"

Post by d14b0ll0s » 19 Aug 2015, 07:41

Tulon wrote:Keep in mind simultaneous processing of multiple pages is only available in 64bit version of ST experimental.
Yes, that's the one I'm using right now. Although there was a 15-20% speed improvement at the final stage (#6: Output), the deskewing (#3) part took a little longer than the speed gain from that, having been moved to Geometric Distortions now. I'll try some more conversions soon with larger samples, but perhaps the speed bump is limited to more powerful machines than my ultrabook (I don't think thermal throttling was the issue, since it is being cooled with a large external fan and doesn't get very warm these days).

xerum
Posts: 41
Joined: 12 Jul 2015, 04:23
Number of books owned: 0
Country: australia

Re: Scan Tailor "Experimental"

Post by xerum » 19 Aug 2015, 08:15

d14b0ll0s wrote: Although there was a 15-20% speed improvement at the final stage (#6: Output), the deskewing (#3) part took a little longer than the speed gain from that,
i assume you went into TOOLS > SETTINGS and checked the two "Accelerate" options.

I'm not sure what CPU or GFX your Ultrabook has but with these two options selected i found to have 500% speed improvement. I am running an i7 CPU. I tried it on my 5 year old Lenovo i3 CPU laptop and i got 200% improvement.

I'm talking final stage taking 20-30 seconds for 150-200 page scans instead of minutes.

There is an issue with Geometric distortion when you select "Curved lines" or "key stones" that can heavily slow down STE which Tulon is in the process of fixing. However if you select "No distortion" or "Page skew" speed performance is not impacted at all.

d14b0ll0s
Posts: 31
Joined: 17 Aug 2015, 19:37
Number of books owned: 3000
Country: Poland

Re: Scan Tailor "Experimental"

Post by d14b0ll0s » 19 Aug 2015, 08:30

xerum wrote:i assume you went into TOOLS > SETTINGS and checked the two "Accelerate" options.
[...]
There is an issue with Geometric distortion when you select "Curved lines" or "key stones" that can heavily slow down STE which Tulon is in the process of fixing. However if you select "No distortion" or "Page skew" speed performance is not impacted at all.
I did. In fact, after a couple of tests I left only the first one checked, since the performance with OpenCL was a little bit slower than without it -- I'm not sure what is the reason and if this is specific to my chip (2-core i7-3537 2.0 GHz + HD 4000 integrated graphics) or software (I'm on Windows 10 since recently, which gives me very mixed feelings). Perhaps the slowness might also have to do with the fact that I'm using 14-16 Mpix JPEG pictures with very little compression as my input? But it might also be as simple as you're saying -- I need to use 'Curved lines' deskewing with all of my portable camera shots of the books, so it might be the primary bottleneck here.

xerum
Posts: 41
Joined: 12 Jul 2015, 04:23
Number of books owned: 0
Country: australia

Re: Scan Tailor "Experimental"

Post by xerum » 19 Aug 2015, 08:52

i'm on 4-core full blown i7-3770
additionally i'm using a GTX465 graphics card
Win7

Now i understand with 2-cores and onboard graphics engine may not be giving the performance hit i am experiencing. Especially with 14mpix photos.

My scans are the standard run of the mill 300 dpi scans using canon desktop scanner.

My eldest son is a full blown gamer and he has one of these graphics cards called a "GTX TITAN". It is worth more than his PC and needs it's own power station to power it. I visited him one day with USB key in hand and I asked him let me use STE on it as i needed to process a 2,154 page scan.

STE with all acceleraters on and the TITAN selected just whipped through 2000+ pages in seconds. Literally a few blinks of the eyes.

This new Tulon development supporting multiprocessor and GFX is a god send. I hope when Tulon fixes the speed issue for curved selection will give you a measurable performance gain.

Tulon
Posts: 687
Joined: 03 Oct 2009, 06:13
Number of books owned: 0
Location: London, UK
Contact:

Re: Scan Tailor "Experimental"

Post by Tulon » 19 Aug 2015, 11:10

d14b0ll0s wrote:2-core i7-3537 2.0 GHz + HD 4000 integrated graphics
OpenCL with i7-3537 may well beat OpenCL with HD 4000. Try that combination. The gain won't be significant though.
Scan Tailor experimental doesn't output 96 DPI images. It's just what your software shows when DPI information is missing. Usually what you get is input DPI times the resolution enhancement factor.

d14b0ll0s
Posts: 31
Joined: 17 Aug 2015, 19:37
Number of books owned: 3000
Country: Poland

Re: Scan Tailor "Experimental"

Post by d14b0ll0s » 19 Aug 2015, 14:18

Tulon wrote:
d14b0ll0s wrote:2-core i7-3537 2.0 GHz + HD 4000 integrated graphics
OpenCL with i7-3537 may well beat OpenCL with HD 4000. Try that combination. The gain won't be significant though.
Thanks for the suggestion. I've tried all the options, and surprisingly the non-OpenCL multi-threaded default one was a bit faster than OpenCL with either the CPU or the GPU.
I've just ordered a 2012 Mac Mini with quad-core i7-3615QM and an SSD -- just waiting for it to come. I'll see if it helps any, provided that I can compile the sources or that running Windows (7?) in Parallels won't be too much of a slowdown.

d14b0ll0s
Posts: 31
Joined: 17 Aug 2015, 19:37
Number of books owned: 3000
Country: Poland

Re: Scan Tailor "Experimental"

Post by d14b0ll0s » 19 Aug 2015, 14:22

xerum wrote:STE with all acceleraters on and the TITAN selected just whipped through 2000+ pages in seconds. Literally a few blinks of the eyes.
That's quite impressive! I know that even lower-cost modern dedicated GPUs can be 10x faster than quad-core i7 chips with lower TDP, but apart from the new options supported by the STE I have very little use for a dedicated graphics card, comparing it against its price, weight, use of power, etc., since I'm mostly trying to stay mobile with my devices.

If, however, any of you has a good idea on an older, cheaper computer (preferrably a laptop or a Mac Mini size computer), I would be happy to consider something with dedicated graphics which would give a speed bump over the 3615QM in the 2012 Mini. Only if you have something in mind though, I don't want anyone to do Googling for me. ;)

dtic
Posts: 464
Joined: 06 Mar 2010, 18:03

Re: Scan Tailor "Experimental"

Post by dtic » 19 Aug 2015, 15:29

xerum wrote:STE with all acceleraters on and the TITAN selected just whipped through 2000+ pages in seconds. Literally a few blinks of the eyes.
Wow, that is really something!
With these speeds STE could, if some other dev figured out how to add command line support, be used for near immediate display of the ST output on a screen during the the scanning process with Spreads or similar tools. Jbaiter and duerig (if you are tracking this thread), any thoughts/plans on that front?

Tulon
Posts: 687
Joined: 03 Oct 2009, 06:13
Number of books owned: 0
Location: London, UK
Contact:

Re: Scan Tailor "Experimental"

Post by Tulon » 19 Aug 2015, 18:23

I just wanted to re-iterate that Scan Tailor experimental doesn't output 96 DPI images. I did mention this twice before. This time I am putting it into my signature.
Scan Tailor experimental doesn't output 96 DPI images. It's just what your software shows when DPI information is missing. Usually what you get is input DPI times the resolution enhancement factor.

Post Reply