Page 1 of 1

opening images directly in CP editor, how?

PostPosted: Tue Jun 29, 2010 7:07 pm
by digipano
Since APG2.5 A1 has manual ability to add CP, is there a way to bypass the detection & optimization & open all the images directly in CP editor to have manual control?

I want to test a difficult gigapixel pano with 1200 images so I dont want to waste time in detection & optimization which will be in few hours & i know it will be wrong due to similar looking images &i will finally need to add manual cp, so I rather add manual CP on images directly in CP editor.

PostPosted: Tue Jun 29, 2010 7:15 pm
by mediavets
Are you really planning to stitch a 1200 image pano using just manually placed CPs?

PostPosted: Tue Jun 29, 2010 7:24 pm
by digipano
Yes its a map screen shot in over 1200 images & will have lot of repetitive feature which will surely confuse APG & after hours it will show me garbled pano I will eventually have to delete all links & try out semi auto cp adding.

In 2.5 you draw rectangle on 1 image then to the 2nd image APP will auto detect the CP in that area, if you don't get good enough points then add few manually. This set of images are best test for this new feature & I want to test this ability.

PostPosted: Tue Jun 29, 2010 7:51 pm
by mediavets
OK, but if you add 1200 images in an undifferentiated 'heap' how will you find near neighbours so you add CPs manually?

PostPosted: Tue Jun 29, 2010 7:57 pm
by digipano
They are already numbered when the screen shots were being taken so I have a series to follow plus the visual clue of the map features.

PostPosted: Tue Jun 29, 2010 8:42 pm
by mediavets
OK how about this wild idea to get your images into APG avoiding detecton and optimisation?

Create a group with just say two of your inages, stitch those and open in the Panorama Editor then in the Layers pane Add the rest of the images.

Will that do it?

PostPosted: Tue Jun 29, 2010 10:14 pm
1200 images need 4800 links --> minimum amount of CPs: 15,000 or 20,000...

Like Mediavets, I have doubts about the manual placement of several thousands of CPs and don't believe Autopano manual CPs were added to replace automatic CP but only to add CPs where SIFT failled to place them correctly.

This being said:

1) You could begin with two images only and then use the Layers Editor to add the other ones (add an image, set its CPs, add an image, set its CPs, etc.)

2) Another possibility (but trying that for 1200 images in a first attempt would be :mad::mad::mad:) is to use a "detection without optimization" parameter setting. For that you must deselect both First optimization and Final optimization in the Optimization settings. CP will be created (you could delete them only where they look not good enough) and you can limit the amount of CPs per link to 10 using the Detection settings.

PostPosted: Wed Jun 30, 2010 4:03 am
by hankkarl
did you use a regular pattern, like left to right, then down a row, etc?

One of the papywizard type imports may help.

PostPosted: Wed Jun 30, 2010 4:13 am
by digipano
this method should work but I am also thinking of another possibility which Gurl suggested.

your method seems better but a slight variation here
add 2 images detect>optimize now add 3-4 & then repeat detect>optimize further the same for all the images & slowly but steadily complete the jigsaw puzzle. Any downside to detect>optimize after every image?

I will try both with a smaller resized set of images.

PostPosted: Wed Jun 30, 2010 6:37 am
by leedsjoe
digipano wrote:Any downside to detect>optimize after every image?

A general point: in my experience, the more individual processes one does in any session the sooner APP/G crashes, most likely because of memory management problems. A partial solution is, of course, to keep saving the project file as one goes. I also use 'save workspace' to assist a restore after crash.

PostPosted: Wed Jun 30, 2010 7:09 am
by digipano
Good point, I should remember saving every 1 min,
auto save feature is a must for such giga projects.

PostPosted: Wed Jun 30, 2010 10:52 am
by Stephmw
Autosave and also some (configurable) saved file history I'd guess too. (gigapano.pano, gigapano.pano.0 ... akin to unix log files)

PostPosted: Wed Jun 30, 2010 12:50 pm
by AlexandreJ
On giga project, you can also rely on the customized level in the undo tool. Put there a low value, so the memory won't raise too much while doing operation in the editor.
BTW : if the project is a gigapan project, just unchecked optimize and you can get a quick overview where the issue are location.
Then play also with the slider in the control point helper windows to hide links which are good ( under a given threshold ). This is nice to detect wrong links.

PostPosted: Wed Jun 30, 2010 2:49 pm
by digipano
Alex this is giga project as per the file size but not shot with any camera, instead these are screen shots from a mapping program. I will follow your advice, planning to test this weekend.

PostPosted: Wed Jun 30, 2010 3:11 pm
by AlexandreJ
Screenshot of a mapping software. Well, then no structure in the shooting :(
So, I would process this way : standard detection, optimization in quick mode, then analysis with the slider.