Page 2 of 2

Re: Conditions for Increased Page Splitting Errors

Posted: 07 Aug 2010, 15:14
by univurshul
If you're experiencing long sessions of manual adjustments like I did, go through the tutorial video a couple times: http://vimeo.com/12524529.

After entering the correct DPIs from my images, the accuracy rate of the page splitting & content selection stabilized.

Re: Conditions for Increased Page Splitting Errors

Posted: 23 Oct 2010, 03:50
by univurshul
Recurring page splitting errors unrelated to DPI input are solved & explained here: http://www.diybookscanner.org/forum/vie ... 3&start=10

Re: Conditions for Increased Page Splitting Errors

Posted: 23 Oct 2010, 13:18
by ibr4him
I have this exact same problem — high error rate with auto split. I'm spending hours adjusting the splits manually too. I really miss the 'Sort by width/height' feature in the 'Split' section.

Also I think it would be a great idea to be able to adjust threshold for page splitting and content selection aggressiveness?

Re: Conditions for Increased Page Splitting Errors

Posted: 23 Oct 2010, 13:43
by Tulon
ibr4him wrote:I have this exact same problem — high error rate with auto split. I'm spending hours adjusting the splits manually too. I really miss the 'Sort by width/height' feature in the 'Split' section.
The solution to this problem is not more options, but adjusting the way you shoot your pages. You don't want the opposite edge of the book to be visible, so you either zoom in until it's not, or move the camera a bit, or as a last resort, do batch cropping using some 3rd party software.
ibr4him wrote:Also I think it would be a great idea to be able to adjust threshold for page splitting and content selection aggressiveness?
Adjusting aggressiveness on the page splitting stage will do no good, as you don't know what's going to be more prominent - the edge or the spine.
Select Content is not a single algorithm, but a combination of many, so there is no single place where aggressiveness can be adjusted.

Also note that I am not going to implement any new features until I am done with dewarping, and I'd rather not receive feature requests until then.

Re: Conditions for Increased Page Splitting Errors

Posted: 23 Oct 2010, 15:23
by univurshul
ibr4him wrote:I have this exact same problem — high error rate with auto split. I'm spending hours adjusting the splits manually too
ibr4him,

I highly suggest you read my replies in this post here: http://www.diybookscanner.org/forum/vie ... ?f=8&t=633

It is a discussion thread about splitting features in ST, but my replies have nothing to do with feature requests; my replies explain why people are getting high errors and how to prevent this from happening & how to correct those books that have already been shot. I show clear evidence to back this up. How you shoot your books with your cameras truly matters. You must read it, it will save you a ton of time.

Don't throw images at Scan Tailor blindly. Practice my guidelines, and you won't waste your time in the future.

Re: Conditions for Increased Page Splitting Errors

Posted: 24 Oct 2010, 06:34
by ibr4him
Yes univurshul, I did try your ideas but the error rate is still too high, especially for this huge book that I'm doing right now.

Anyway, I think a "snap to (content?)" feature would be incredibly useful in ST, it should just like Photoshop's 'snap to guides' feature. I understand the algorithm can't be 100% accurate so this feature will be a huge time-saver for manual edits. Hope Tulon thinks the same! :)

Re: Conditions for Increased Page Splitting Errors

Posted: 24 Oct 2010, 06:39
by univurshul
ibr4him wrote:Yes univurshul, I did try your ideas but the error rate is still too high
Post an image or load several onto a dropbox account and send me the public URL. I will be glad to show you where it's causing errors and how to prevent it.

Within exception to an occasional column split error, I have nearly zero split errors. It is ridiculously low.