Panorama detection taking a lot of time  

Share your tips and tricks here or get help with any Autopano Pro / Giga problem!
No bug reports (of any kind) in this forum!
no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

Panorama detection taking a lot of time

by hermer-blr » Fri May 11, 2012 12:27 pm

For the time I have been using APG 2.6 (maybe it started earlier...), I have noticed that the time taken to perform a panorama detection is very very very long. My memory from the past is that it took some tens of seconds, while now it easily takes 10 minutes ! My most recent panoramas only comprised 8 fisheye images; I would have expected the detection to be quick (compared to my multiple exposure 24 x 3 images panoramas), but nothing was different...

Is there anything wrong with my configuration (or my installation of APG 2.6) or is this time normal ? I must say that I check all the optimization parameters (multiple viewpoint, final optimization, strong algorithm, etc...) with the hope that it will give a perfect detection and panorama assembly.

Note: When, after having changed, the control points, I launch a full optimization, it is then very fast
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Fri May 11, 2012 4:27 pm

Detection times can vary. The following cause can change the total time :
- By far, RAW decoding can really slow down
- Formats : jpeg is faster than using 16bits TIFF ( because of image format size )
- Optimizer options : fisheye is the longest
- the number of images : the more the longer ( obvious, but less quote that ).

no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

by hermer-blr » Fri May 11, 2012 5:57 pm

I always perform raw to tif conversion prior to using APG. My images are 16 bit tif images.

But 10 mn for the detection of a panorama with 8 fisheye images, is that normal ??? My configuration is 3 years old, with 4 Gbytes, dual core processor...
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

by hermer-blr » Fri May 11, 2012 8:38 pm

To be accurate:

Input: 8 16 bit tif files x 16 MPixels - Samyang fisheye

Time taken for panorama detection: 28 minutes !!! Most of the time was taken by "control point validation" - Note that, on my computer, the panorama stitching to 12000 x 6000 points takes about 30 minutes (not really more than the detection).

My configuration is Windows 32 bits, 4 Gbytes, running on a dual core AMD 5000+ (2.61 GHz)

Result : bad RMS of 3.38

Furthermore, even when optimizing the control points and reaching an excellent RMS (1.7), I could notice that the average quality of the panorama was not really good:
- I also requested the export of the images for post-processing purpose
- The first image that I opened was strongly distorted when superimposed to the panorama
So, the panorama was correct only because of the deghosting, able to hide most of the features misalignments

I am really perplex !!!!
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

by hermer-blr » Fri May 11, 2012 9:36 pm

New test made by changing both detection quality from high to standard and ransac model from homography to simiarity; then, the panorama detection is very fast !

I however confirm after having performed an other panorama optimization that the distortion correction is rather bad. When I edit the panorama and put my mouse on the different images, I can see the important fluctuation in the features ! So, exporting the images for post-processing is useless !
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

by hermer-blr » Fri May 11, 2012 10:45 pm

Hum !

I must confess that without a solution to efficiently render my panorama I gave a try to PTGUI pro.

Without any knowledge of the software and without any previous experience of it, I could easily stitch my panorama perfectly and using the mask function perform all the maskings I wanted without any need for further post-processing.

I can thus derive from this experience that my 8 shots were correct and that there is really a problem with APG unable to make a proper distortion correction of my fisheye lens. The other lesson lernt is that I am tempted to buy PTGUI Pro... so easy it is to use !
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

no avatar
mediavets
Moderator
 
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

by mediavets » Fri May 11, 2012 11:11 pm

hermer-blr wrote:Is there anything wrong with my configuration (or my installation of APG 2.6) or is this time normal ? I must say that I check all the optimization parameters (multiple viewpoint, final optimization, strong algorithm, etc...) with the hope that it will give a perfect detection and panorama assembly.

If you are using a fisheye lens on a proper pano head why are you selecting multiple viewpoint correction?
Andrew Stephens
Many different Nodal Ninja and Agnos pano heads. Merlin/Panogear mount with Papywizard on Nokia Internet tablets.
Nikon D5100 and D40, Sigma 8mm f3.5 FE, Nikon 10.5mm FE, 35mm, 50mm, 18-55mm, 70-210mm. Promote control.

no avatar
mediavets
Moderator
 
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

by mediavets » Fri May 11, 2012 11:23 pm

hermer-blr wrote:I always perform raw to tif conversion prior to using APG. My images are 16 bit tif images.

Is your Samyang the AE model for Nikon?

What focal length do you specify is you are entering the lens type and focal length manually?
But 10 mn for the detection of a panorama with 8 fisheye images, is that normal ??? My configuration is 3 years old, with 4 Gbytes, dual core processor...

I am also still running Windows XP-32 bit OS and my impression if that for APP/APG 2.6x the software really demands a 64-bit OS with more than 4GB of RAM.

Image sets that I could process easily with 2.5x I now really struggle to handle with 2.6x.

Do you have a supported GPU-based graphics card?
Andrew Stephens
Many different Nodal Ninja and Agnos pano heads. Merlin/Panogear mount with Papywizard on Nokia Internet tablets.
Nikon D5100 and D40, Sigma 8mm f3.5 FE, Nikon 10.5mm FE, 35mm, 50mm, 18-55mm, 70-210mm. Promote control.

User avatar
gkaefer
Member
 
Posts: 3549
Likes: 0 post
Liked in: 15 posts
Joined: Tue Jun 09, 2009 1:01 pm
Location: Salzburg
Info

by gkaefer » Fri May 11, 2012 11:40 pm

samyang 8mm is a manual lens. did you set lens, focal value and crop factor values inside autopano image properties to fisheye?
samyang 8mm ... I allway did use 8mm in the settings.
but the lens is available from many brands, some saying its 7,8,10mm... and on autopano forum I read most fitting should be 7mm..
win32 with 4gig ram... was also my last non win7pc I had. and at that time the 32bit version of autopanowas limited by the hardware resouces. I had many crashes and pano with 100images were the limit and it took decades of hours.... same project wit 8gig Win7 64bit and 64bit autopano did speed up the system (down to minutes) like a saturn rocket leaving the ramp. so if you plan to do larger panos with 100-400 images than I would strong give you the tip to change to 64bit system.

also the samyang 8mm has a golden ring. about 1mm from this ring in direction to camera the NPP is placed. I had specially indoors stitching problems if the NPP is here wrong placed only 1mm.

Liebe Gruesse,
Georg

no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

by hermer-blr » Sat May 12, 2012 1:13 pm

To mediavets :

1/ Multiple viewpoint correction: maybe because I am not so sure of my head calibration that I have been using for the first time last week during a vacation trip in New-York.

Old habit also, because until now (before the Samyang fisheye) I was shooting all my panoramas handheld (however "handheld" trying to take care of the NPP).

2/ Yes, my Samyang is the AE. As it was not in the lens database, I have added it myself, putting the focal length to 8 mm. The focal length given by APG after processing is 7.83 (after diivision by 1.558). The focal length given by PTGui after processing is 8.76 mm. The PTGUI stitching result is perfect.

3/ Yes I have GPU based graphic card. But see, my other reply here : By changing the APG parameters, I could put the time from 28 mn (it was actually much more than 10 mn) to about 1 mn.
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

no avatar
hermer-blr
Member
 
Topic author
Posts: 237
Likes: 0 post
Liked in: 0 post
Joined: Thu May 06, 2010 1:01 pm
Location: Near Paris - France
Info

by hermer-blr » Sat May 12, 2012 1:22 pm

To gkaefer:

1/ I used 8 mm (see answer to Mediavets).

2/ Regarding the location of the NPP, my reference is here :

http://www.hugha.co.uk/NodalPoint/Index.htm

As you can see, it is a few millimeters from the golden ring, towards the lens surface, ie opposite to the direction you propose.

I should verify it by myself...

3/ As mentioned in this post, PTGUI perfectly stitches the panorama; it calculates a focal length of 8.76 mm...
Nikon D5100 (formerly Nikon D60) - Sigma 10-20 - 24 shots Panoramas in 3 raws
Windows 32 bits - APG
website htt://jmh.trp.free.fr

no avatar
mediavets
Moderator
 
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

by mediavets » Sat May 12, 2012 2:13 pm

hermer-blr wrote:2/ Regarding the location of the NPP, my reference is here :

http://www.hugha.co.uk/NodalPoint/Index.htm

As you can see, it is a few millimeters from the golden ring, towards the lens surface, ie opposite to the direction you propose.

I should verify it by myself...

Yes, you should.

The positioning on the lower rail is more important that positioning on the upper rail - Hans Nyberg has demonstrated this.

This may help test the lower rail position:
http://www.easypano.com/forum/display_topic_threads.asp?ForumID=1&TopicID=4162

It is also very important that the plane of the camera sensor is perpendicular to the upper rail.
Andrew Stephens
Many different Nodal Ninja and Agnos pano heads. Merlin/Panogear mount with Papywizard on Nokia Internet tablets.
Nikon D5100 and D40, Sigma 8mm f3.5 FE, Nikon 10.5mm FE, 35mm, 50mm, 18-55mm, 70-210mm. Promote control.


Who is online

Users browsing this forum: No registered users and 4 guests