cpfind memory allocation error

[Brent Townshend]

On running cpfind with 12 16-bit TIFF files, each 5616x3744, cpfind is crashing as follows. Perhaps it should have an option to downsample further or do so automatically. I'm running on a mac with 32GBytes of RAM.

ThreadQueue created
Hugins cpfind 2010.4.0 built by Harry van der Wolf
based on Pan-o-matic by Anael Orlinski

Input file : /var/folders/8h/8hlpRcCCEaGFmRb8QB9D-E+++TI/-Tmp-/ap_inprojFJqZwc
Output file : /var/folders/8h/8hlpRcCCEaGFmRb8QB9D-E+++TI/-Tmp-/ap_resTLTV9P
Number of CPU : 8

Input image options
  Downscale to half-size : yes
Gradient based description
Sieve 1 Options
  Width : 10
  Height : 10
  Size : 50
  ==> Maximum keypoints per image : 5000
KDTree Options
  Search steps : 40
  Second match distance : 0.15
Matching Options
  Mode : All pairs
  Distance threshold : 25
RANSAC Options
  Iterations : 1000
  Distance threshold : 25
Sieve 2 Options
  Width : 5
  Height : 5
  Size : 2
  ==> Maximum matches per image pair : 50
User thread created.
Reference thread created.
1 reference-thread added.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...
User thread created.
pollPendingThreads()
1 user-thread added.
Thread starting...

Project contains the following images:
Image 0
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0182.tif
  Remapped : yes
Image 1
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0183.tif
  Remapped : yes
Image 2
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0185.tif
  Remapped : yes
Image 3
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0186.tif
  Remapped : yes
Image 4
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0187.tif
  Remapped : yes
Image 5
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0188.tif
  Remapped : yes
Image 6
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0189.tif
  Remapped : yes
Image 7
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0190.tif
  Remapped : yes
Image 8
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0191.tif
  Remapped : yes
Image 9
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0192.tif
  Remapped : yes
Image 10
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0184.tif
  Remapped : yes
Image 11
  Imagefile: /Volumes/Adams/Users/bst/Desktop/Hugin Export/_MG_0193.tif
  Remapped : yes

--- Analyze Images ---
i0 : Load image...
i1 : Load image...
i2 : Load image...
i3 : Load image...
i4 : Load image...
i5 : Load image...
i6 : Load image...
i7 : Load image...
i5 : Remap image...
i6 : Remap image...
i1 : Remap image...
i3 : Remap image...
i7 : Remap image...
i4 : Remap image...
i2 : Remap image...
i0 : Remap image...
i2 : Convert to greyscale double...
i5 : Convert to greyscale double...
i3 : Convert to greyscale double...
i2 : Build integral image...
i5 : Build integral image...
i3 : Build integral image...
i2 : Build distance map...
i3 : Build distance map...
i4 : Convert to greyscale double...
i5 : Build distance map...
i4 : Build integral image...
i4 : Build distance map...
i6 : Convert to greyscale double...
i2 : Find keypoints...
i3 : Find keypoints...
i5 : Find keypoints...
i6 : Build integral image...
i7 : Convert to greyscale double...
i6 : Build distance map...
i7 : Build integral image...
i4 : Find keypoints...
i7 : Build distance map...
i0 : Convert to greyscale double...
i1 : Convert to greyscale double...
i0 : Build integral image...
i1 : Build integral image...
i6 : Find keypoints...
i0 : Build distance map...
i1 : Build distance map...
i7 : Find keypoints...
i0 : Find keypoints...
i1 : Find keypoints...
i5 : Found 42196 interest points.
i5 : Filtering keypoints...
i5 : Kept 2400 interest points.
i5 : Make keypoint descriptors...
i5 : Build KDTree...
i5 : Freeing memory...
i8 : Load image...
i3 : Found 47339 interest points.
i3 : Filtering keypoints...
i3 : Kept 2400 interest points.
i3 : Make keypoint descriptors...
i3 : Build KDTree...
i3 : Freeing memory...
i9 : Load image...
An error happened while loading image : caught exception: std::bad_alloc
i10 : Load image...
An error happened while loading image : caught exception: std::bad_alloc
i11 : Load image...
An error happened while loading image : caught exception: std::bad_alloc
cpfind(68746,0xb0081000) malloc: *** mmap(size=252317696) failed (error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
cpfind(68746,0xb0081000) malloc: *** mmap(size=252317696) failed (error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
cpfind(68746,0xb0081000) malloc: *** mmap(size=252317696) failed (error code=12)
*** error: can't allocate region
*** set a breakpoint in malloc_error_break to debug
i2 : Found 40564 interest points.
i2 : Filtering keypoints...
i2 : Kept 2398 interest points.
i2 : Make keypoint descriptors...
i2 : Build KDTree...
i2 : Freeing memory...
i8 : Remap image...
i4 : Found 45658 interest points.
i4 : Filtering keypoints...
i4 : Kept 2400 interest points.
i4 : Make keypoint descriptors...
i4 : Build KDTree...
i4 : Freeing memory...
i1 : Found 47289 interest points.
i1 : Filtering keypoints...
i6 : Found 49432 interest points.
i6 : Filtering keypoints...
i7 : Found 49374 interest points.
i7 : Filtering keypoints...
i6 : Kept 2358 interest points.
i6 : Make keypoint descriptors...
i1 : Kept 2400 interest points.
i1 : Make keypoint descriptors...
i0 : Found 44658 interest points.
i0 : Filtering keypoints...
i7 : Kept 2350 interest points.
i7 : Make keypoint descriptors...
i0 : Kept 2400 interest points.
i0 : Make keypoint descriptors...
i6 : Build KDTree...
i1 : Build KDTree...
i6 : Freeing memory...
i7 : Build KDTree...
i1 : Freeing memory...
i7 : Freeing memory...
i0 : Build KDTree...
i0 : Freeing memory...

[Yuv (yuv)]

It seems that you are running 2010.4.0. Cpfind has been massively improved since. Can you download a recent snapshot of 2010.5.0 from http://panorama.dyndns.org/index.php?lang=en&subject=hugin&texttag=hugin and try again? Please set the status of this report to New when adding new information. Thanks.

[Brent Townshend (bst)]

The bug still occurs with this snapshot, see attached command output.

[Yuv (yuv)]

not sure if it is the same error. 8GB RAM Kubuntu with 293 8 megapixel JPEGs fails on image 15

cpfind -v -o out.pto --multirow 090601nendaz01.input.pto

....
Image 292
  Imagefile: _MG_8039_01.jpg
  Remapped : no
Image 293
  Imagefile: _MG_8040_01.jpg
  Remapped : no

--- Analyze Images ---
i0 : Load image...
i1 : Load image...
i0 : Resize to greyscale double...
i1 : Resize to greyscale double...
i0 : Build integral image...
i0 : Find keypoints...
i1 : Build integral image...
i1 : Find keypoints...
i1 : Found 30643 interest points.
i1 : Filtering keypoints...
i1 : Kept 10000 interest points.
i1 : Make keypoint descriptors...
i0 : Found 22730 interest points.
i0 : Filtering keypoints...
i0 : Kept 9600 interest points.
i0 : Make keypoint descriptors...
i1 : Build KDTree...
i0 : Build KDTree...
i1 : Freeing memory...
i2 : Load image...
i0 : Freeing memory...
i3 : Load image...
i2 : Resize to greyscale double...
i3 : Resize to greyscale double...
i2 : Build integral image...
i2 : Find keypoints...
i3 : Build integral image...
i3 : Find keypoints...
i2 : Found 19085 interest points.
i2 : Filtering keypoints...
i2 : Kept 7960 interest points.
i2 : Make keypoint descriptors...
i3 : Found 10966 interest points.
i3 : Filtering keypoints...
i3 : Kept 5688 interest points.
i3 : Make keypoint descriptors...
i3 : Build KDTree...
i3 : Freeing memory...
i4 : Load image...
i2 : Build KDTree...
i2 : Freeing memory...
i5 : Load image...
i4 : Resize to greyscale double...
i5 : Resize to greyscale double...
i4 : Build integral image...
i5 : Build integral image...
i4 : Find keypoints...
i5 : Find keypoints...
i4 : Found 11369 interest points.
i4 : Filtering keypoints...
i4 : Kept 5199 interest points.
i4 : Make keypoint descriptors...
i5 : Found 10729 interest points.
i5 : Filtering keypoints...
i5 : Kept 5544 interest points.
i5 : Make keypoint descriptors...
i5 : Build KDTree...
i5 : Freeing memory...
i6 : Load image...
i4 : Build KDTree...
i4 : Freeing memory...
i7 : Load image...
i6 : Resize to greyscale double...
i7 : Resize to greyscale double...
i6 : Build integral image...
i6 : Find keypoints...
i7 : Build integral image...
i7 : Find keypoints...
i6 : Found 11600 interest points.
i6 : Filtering keypoints...
i6 : Kept 6404 interest points.
i6 : Make keypoint descriptors...
i7 : Found 9610 interest points.
i7 : Filtering keypoints...
i7 : Kept 3941 interest points.
i7 : Make keypoint descriptors...
i7 : Build KDTree...
i7 : Freeing memory...
i8 : Load image...
i6 : Build KDTree...
i6 : Freeing memory...
i9 : Load image...
i8 : Resize to greyscale double...
i8 : Build integral image...
i9 : Resize to greyscale double...
i8 : Find keypoints...
i9 : Build integral image...
i9 : Find keypoints...
i8 : Found 9560 interest points.
i8 : Filtering keypoints...
i8 : Kept 3065 interest points.
i8 : Make keypoint descriptors...
i8 : Build KDTree...
i8 : Freeing memory...
i10 : Load image...
i9 : Found 9513 interest points.
i9 : Filtering keypoints...
i9 : Kept 6772 interest points.
i9 : Make keypoint descriptors...
i9 : Build KDTree...
i10 : Resize to greyscale double...
i9 : Freeing memory...
i11 : Load image...
i10 : Build integral image...
i10 : Find keypoints.....

[Yuv (yuv)]

when running in gdb, I get a little extra info:

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffeb60c700 (LWP 29440)]
0x00000000004690e7 in flann::KDTreeIndex<flann::L2<double> >::divideTree(int*, int) ()

[Bob Bright (bbbright)]

I'm unable to confirm this bug on Xubuntu 10.04. I just tried running cpfind from the command line with default settings on a set of 15 16-bit .tifs with a resolution of 3745x5626 (generated from 5DMkII .CR2 files, as I presume Brent's images were). Cpfind took quite a while (11 minutes) but ran to completion.

This was on an aging Intel Core2 Duo laptop with a meager 2GB of RAM, so whatever problem Brent is experiencing, it doesn't seem likely that further downscaling of the images is going to solve it.

Cheers,
BBB

bbb@BBB3:~/Pictures/5DMkII/11.03.05/test4-10mm$ cpfind -v -o out.pto new.pto
Hugins cpfind 2011.0.0.8f1447ab8649
based on Pan-o-matic by Anael Orlinski
Input file : new.pto
Output file : out.pto
Number of CPU : 2

Input image options
  Downscale to half-size : yes
Gradient based description
Sieve 1 Options
  Width : 10
  Height : 10
  Size : 100
  ==> Maximum keypoints per image : 10000
KDTree Options
  Search steps : 200
  Second match distance : 0.25
Matching Options
  Mode : All pairs
  Distance threshold : 50
RANSAC Options
  Mode : Iterations : 1000
  Distance threshold : 50
Sieve 2 Options
  Width : 5
  Height : 5
  Size : 1
  ==> Maximum matches per image pair : 25

Project contains the following images:
Image 0
  Imagefile: IMG_5747.tif
  Remapped : yes
Image 1
  Imagefile: IMG_5748.tif
  Remapped : yes
Image 2
  Imagefile: IMG_5749.tif
  Remapped : yes
Image 3
  Imagefile: IMG_5750.tif
  Remapped : yes
Image 4
  Imagefile: IMG_5751.tif
  Remapped : yes
Image 5
  Imagefile: IMG_5752.tif
  Remapped : yes
Image 6
  Imagefile: IMG_5753.tif
  Remapped : yes
Image 7
  Imagefile: IMG_5754.tif
  Remapped : yes
Image 8
  Imagefile: IMG_5755.tif
  Remapped : yes
Image 9
  Imagefile: IMG_5756.tif
  Remapped : yes
Image 10
  Imagefile: IMG_5757.tif
  Remapped : yes
Image 11
  Imagefile: IMG_5758.tif
  Remapped : yes
Image 12
  Imagefile: IMG_5759.tif
  Remapped : yes
Image 13
  Imagefile: IMG_5760.tif
  Remapped : yes
Image 14
  Imagefile: IMG_5761.tif
  Remapped : yes

--- Analyze Images ---
i0 : Load image...
i1 : Load image...
i0 : Remap image...
i1 : Remap image...
i0 : Convert to greyscale double...
i0 : Build integral image...
i0 : Build distance map...
i1 : Convert to greyscale double...
i1 : Build integral image...
i1 : Build distance map...
i0 : Find keypoints...
i1 : Find keypoints...
i0 : Found 156181 interest points.
i0 : Filtering keypoints...
i0 : Kept 8616 interest points.
i0 : Make keypoint descriptors...
i0 : Remapping back keypoints...
i0 : Build KDTree...
i0 : Freeing memory...
i2 : Load image...
i1 : Found 117314 interest points.
i1 : Filtering keypoints...
i1 : Kept 8496 interest points.
i1 : Make keypoint descriptors...
i1 : Remapping back keypoints...
i1 : Build KDTree...
i1 : Freeing memory...
i3 : Load image...
i3 : Remap image...
i2 : Remap image...
i3 : Convert to greyscale double...
i3 : Build integral image...
i3 : Build distance map...
i2 : Convert to greyscale double...
i2 : Build integral image...
i3 : Find keypoints...
i2 : Build distance map...
i2 : Find keypoints...
i3 : Found 150507 interest points.
i3 : Filtering keypoints...
i3 : Kept 8870 interest points.
i3 : Make keypoi...

[tmodes (tmodes)]

The problem is memory usage and number of threads (aka cores).
Looking at the inital report it tries to load and remap 8 images at one time. This steps needs propably "too much" memory.
So we need to reduce memory usage or/and automatically restrict the number of threads depending on image sizes.

Workaround could be to limit the number of threads with command line option --ncores.

[Brent Townshend (bst)]

Using --ncores=1 does indeed workaround the problem, but it shouldn't be necessary...

[the_mechanical (mechanical)]

Same problem for me.
cpfind -o Pano.pto Pano.pto --maxmatches 10 did it, but when i try to use celeste or multirow (or both) then it fails.

cpfind --celeste -v -o Pano.pto Pano.pto --maxmatches 10
...
i97 : Convert to greyscale double...
i97 : Build integral image...
i97 : Build distance map...
i100 : Mask areas with clouds...
i97 : Find keypoints...
i97 : Found 433 interest points.
i97 : Filtering keypoints...
i97 : Kept 0 interest points.
i97 : Make keypoint descriptors...
i97 : Build KDTree...
Speicherzugriffsfehler

Without celeste it works!
cpfind --multirow -v -o Pano.pto Pano.pto --maxmatches 10
..
i204 <> i260 : Kept 11 matches.

--- Write Project output ---
Written output to Pano_Granatenkogel.pto

Detection took 297.058 seconds.

Similar with
cpfind -o Pano.pto Pano.pto --maxmatches 10

System: Ubuntu Natty 64bit, 16GB Memory, 30GB Swap; Phenom II X4 955;
Hugins cpfind Pre-Release 2011.1.0.83a530eea793

[the_mechanical (mechanical)]

Mistake: should be
but when i try to use celeste (alone or with multirow) then it fails.

[Yuv (yuv)]

I can now confirm that what I saw in comment #3 is not related to this thread. My error was due to an image being in the wrong format. It also caused bug 779563. I will file e separate ticket for that. Apology for the thread hijack.

Agree that manually limiting the number of cores should not be necessary. The problem will affect an increasing number of users in the future.

The solution is some logic before spawning the thread. Spawn it only if enough memory can be allocated.

Question: memory = physical RAM? or memory = available memory including (slow) swap?

[Rich (rpow5152)]

+1
I see the same error on Windows XP 64 bit. System is an i5 with 8GB of RAM and 2TB of storage free.
Problem either happens when I find control points on any pair of 16 bit TIF images that are 5616x3744 dimensions (20MP).
If I use multiple 8bit TIF images at 5616x3744 dimension, I also run into problems errors that result in few or no control points found.
If I use 90 8bit TiF images at smaller dimensions (e.g. 7MP) the same problem occurs.

I have spent days manually setting control points on 360 panos. Would be awesome to find a solution to this.

If I upgrade my RAM to 16GB am I likely to overcome this limitation?

[Kurt Hillig (kwhillig)]

FWIW the Autopano-SIFT-C always seems to work - though as far as I can tell
it does a poor job of picking "good" CPs,
in most cases they all seem to be very close together and often on
low-to-zero contrast "features" such as sky,
clouds or deep shadow. And of course there are the patent/license issues
with it...

I've recently upgraded my Mac from 12 to 16 GB, but that hasn't seemed to
help. I generally use 16-bit TIFF
images 5616x3744 - same as yours (also a Canon 5DmkII?) - but as far as I
can tell the determining factor is
the number of images that I try to generate CPs for at one time. I just ran
a test on a pan that fails when I try
to generate CPs for all 15 images at once; but if I select smaller subsets
of these I can generate CPs successfully.

On Thu, Jun 23, 2011 at 9:19 AM, Rich <email address hidden> wrote:

> +1
> I see the same error on Windows XP 64 bit. System is an i5 with 8GB of RAM
> and 2TB of storage free.
> Problem either happens when I find control points on any pair of 16 bit TIF
> images that are 5616x3744 dimensions (20MP).
> If I use multiple 8bit TIF images at 5616x3744 dimension, I also run into
> problems errors that result in few or no control points found.
> If I use 90 8bit TiF images at smaller dimensions (e.g. 7MP) the same
> problem occurs.
>
> I have spent days manually setting control points on 360 panos. Would be
> awesome to find a solution to this.
>
> If I upgrade my RAM to 16GB am I likely to overcome this limitation?
>
> --
> You received this bug notification because you are subscribed to a
> duplicate bug report (789938).
> https://bugs.launchpad.net/bugs/717944
>
> Title:
> cpfind memory allocation error
>
> Status in Hugin - Panorama Tools GUI:
> Confirmed
>
> Bug description:
> On running cpfind with 12 16-bit TIFF files, each 5616x3744, cpfind is
> crashing as follows. Perhaps it should have an option to downsample
> further or do so automatically. I'm running on a mac with 32GBytes
> of RAM.
>
>
> ThreadQueue created
> Hugins cpfind 2010.4.0 built by Harry van der Wolf
> based on Pan-o-matic by Anael Orlinski
>
> Input file :
> /var/folders/8h/8hlpRcCCEaGFmRb8QB9D-E+++TI/-Tmp-/ap_inprojFJqZwc
> Output file :
> /var/folders/8h/8hlpRcCCEaGFmRb8QB9D-E+++TI/-Tmp-/ap_resTLTV9P
> Number of CPU : 8
>
> Input image options
> Downscale to half-size : yes
> Gradient based description
> Sieve 1 Options
> Width : 10
> Height : 10
> Size : 50
> ==> Maximum keypoints per image : 5000
> KDTree Options
> Search steps : 40
> Second match distance : 0.15
> Matching Options
> Mode : All pairs
> Distance threshold : 25
> RANSAC Options
> Iterations : 1000
> Distance threshold : 25
> Sieve 2 Options
> Width : 5
> Height : 5
> Size : 2
> ==> Maximum matches per image pair : 50
> User thread created.
> Reference thread created.
> 1 reference-thread added.
> pollPendingThreads()
> 1 user-thread added.
> Thread starting...
> User thread created.
> pollPendingThreads()
> 1 user-thread added.
> Thread starting...
> User thread created.
> pollPendingThreads()
> 1 user-thread added.
> Thread starting...
> User thr...

[hvdwolf (hvdwolf)]

I can confirm this behavior of cpfind on Mac, for the 2010.4 as well as newer builds. The cpfind in 2010.4 is only 32bits. There is no 64bits version as the 64bit versions segfaulted on startup.
Quite a few improvements have been added to cpfind and I built a 64bit cpfind again, but it's not in any Hugin bundle yet.
However, when trying to use the 64bit version it never works. Not on small images and neither on big/huge images (currently a set of 13678x9394 images).
The 64bit cpfind always segfaults, no longer on startup, but now during processing the images. I only choose 2 images to make it as simple as possible.

Some remarks:
- the 32bit binaries are always built to be able to run on Tiger (OSx 10.4). Tiger was known to have a "weak" implementation of malloc and mmap. The 64bit version is built to run on Leopard (10.5) and newer as Tiger doesn't support 64bit. Leopard and newer are supposed to have ccorrect implementations of malloc and mmap.
- cpfind versions are still built against the internal Zthread library.

I have added a crash log from the 64bit cpfind. I can't deduct anything from it. Hopefully you can.

[hvdwolf (hvdwolf)]

The crash log as described in my previous post. I assumed I could "connect" it to that post but that's not possible.

[Rich (rpow5152)]

When I add --ncores=1 as an argument for cpfind via preferences in Hugin I get :
failed with error code: 1
whenever I try to align with cpfind

Syntax used was: --multirow --ncores=1 -o %o %s
and also --multirow -o %o %s --ncores=1
and --linearmatch --ncores=1 -o %o %s
and --linearmatch -o %o %s --ncores=1

Running on windows 7

[Kay Roepke (kay-roepke)]

Rich, don't use a '=' after --ncores, instead do:
--ncores 1

[Rich (rpow5152)]

Thanks. This worked great.
I now have two big panos with over 1000 good control points. This has saved me hours of work!
CPfind quality is impressive.

[tmodes (tmodes)]

Automatically limit number of threads depending on memory and image size implemented in changeset 6a085a22a136

[Rich (rpow5152)]

Has this fix been released into any Hugin builds yet?
Or should I get the latest cpfind build and replace bin\cpfind.exe ?

[Rich (rpow5152)]

Bump. Is the fixed version available in any of the downloadable binaries yet?

[Max (mrbbbump)]

I am getting this comment while aligning my images in Hugin.

An error happened while loading image : caught exception:

Than it is asking me to select my own control points. This has never happened before and have done 80 stitches in Hugin so far. I have recently changed my camera to a Canon 600d and the files are RAW files converted to JPEG.

I have tried uninstalling and reinstalling the latest version of Hugin but was no use.

Can anyone help???

[Thomas Baeckeroot (thomas.baeckeroot)]

I had this same issue (or similar std::bad_alloc ) with a recent version of cpfind (nightly): 2012.1.0.d604a77b6e02

Here is the outpout:

Hugin's cpfind 2012.1.0.d604a77b6e02
based on Pan-o-matic by Anael Orlinski

Project contains the following images:
Image 0
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6483.JPG
  Remapped : yes
Image 1
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6484.JPG
  Remapped : yes
Image 2
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6485.JPG
  Remapped : yes
Image 3
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6486.JPG
  Remapped : yes
Image 4
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6487.JPG
  Remapped : yes
Image 5
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6488.JPG
  Remapped : yes
Image 6
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6489.JPG
  Remapped : yes
Image 7
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6490.JPG
  Remapped : yes
Image 8
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6491.JPG
  Remapped : yes
Image 9
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6492.JPG
  Remapped : yes
Image 10
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6493.JPG
  Remapped : yes
Image 11
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6494.JPG
  Remapped : yes
Image 12
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6495.JPG
  Remapped : yes
Image 13
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6496.JPG
  Remapped : yes
Image 14
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6497.JPG
  Remapped : yes
Image 15
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6498.JPG
  Remapped : yes
Image 16
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6499.JPG
  Remapped : yes
Image 17
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6500.JPG
  Remapped : yes
Image 18
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6501.JPG
  Remapped : yes
Image 19
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6502.JPG
  Remapped : yes
Image 20
  Imagefile: /home/thomas/Photos/2012-07-30 - Montpellier-le-Vieux (panos à faire)/IMG_6483-6510_pano/IMG_6503.JPG
  Remapped ...