First build with automatic dewarping

[Hamid]

As for me, Beta8 seems to be stable: batch processing with automatic dewarping doesn't cause the programme to crash anymore. I would have never thought it was due to the absence of content boxes.Thanks Tulon.

[emmerick]

Here comes beta8: http://depositfiles.com/files/96wxqhos6
No changes compared to beta7, except a fixed crash on pages without a content box.

Thanks a lot man you are the best.

[dtic]

I got a crash with beta8 when autodewarping ~100 pages on a testrun, but missed to submit the error log I'm afraid. Will submit if I can recreate it. A question: the autodewarping takes significantly longer time compared to the other steps. Is that just how it is for this type of processing or do you think it is possible to later get gains in speed once all else (crashing etc) is worked out? I need to know if I should start saving to buy on of the new Sandy Bridge CPUs

[Tulon]

I got a crash with beta8 when autodewarping

I am going to need the page causing the crash.

A question: the autodewarping takes significantly longer time compared to the other steps. Is that just how it is for this type of processing or do you think it is possible to later get gains in speed once all else (crashing etc) is worked out?

I don't know. I haven't looked into the performance side yet.

[Anonymous1]

@dtic, I profiled the whole program, and the function which takes up the most processing time (by far) is the content-box finding. If I could fluently read C++, I would attempt to see how it could be sped up (I still vote for GPU acceleration ). But for now, Python it is.

[Tulon]

May I remind that I still need the source images for reproducing a crash with beta8. I've got 4 crash reports so far, but I can't reproduce this issue without the source images.

[zamacam]

I've got a crash using Scantailor 1.0 Beta 8. Here the output available in terminal:

imac:~ david$ ./Desktop/scantailor/_scantailor-1.0-beta8/scantailor
Floating point exception
imac:~ david$

Scantailor was running on Mac OS 10.6.6. I builded it from sources (using up to date macport). It worked nicely on several books since Beta 8 is available. It is the first Scantailor crash I've got.

The crash doesn't happen if I disable automatic dewarping.

I'm attaching the faulty image and the file project.

[edit]: the faulty image is image #290 in the project.

[Tulon]

I've reproduced the crash using zamacam's image and project file. That revealed not just one, but two bugs. I'll fix them soon.

[emmerick]

I've reproduced the crash using zamacam's image and project file. That revealed not just one, but two bugs. I'll fix them soon.

Hopefully the next beta always eager ))))

[jimboh]

I see 5 new items committed to the git repository within the last few hours. Are they all dewarping stuff? I don't know the "innards" well enough to know for sure.

One hour later: I downloaded the snapshot as of Sat, 29 Jan 2011 22:30:41 +0000, compiled it with no problems, and obtained the following results with automatic dewarping on a page that was not warped to begin with. The automatic dewarping process warped an unwarped page. I suppose we can just turn dewarping off for pages that are okay, but ... perhaps this shows an area needing improvement. FWIW.

This is the input file:

input.tif

(6.79 MiB) Not downloaded yet

This shows the dewarping "grid" (if that is the correct word):

grid.png (794.34 KiB) Viewed 996 times

This is the output: