Messages

1

Face and Body Scanning / Re: Head scans in the Los Angeles area?

« on: May 08, 2014, 02:35:47 PM »

If your looking for high-end solution, can't beat the ICT Lightstage... google it, but prepare to spend some $$$$

2

General / Re: Yet again out of Memory

« on: July 27, 2012, 10:45:31 PM »

I as well get frequent out of memory errors, with 196 gigs of ram..

Here's a few suggestions based on the 0.9.0 version..

the issue I have, is since the longest part of the process is the depth generation, it's great that now you can re-use depth data, but the problem is you CAN'T re-use it if you change reconstruction resolution, for example. I do a Ultra High reconstruction, takes 2 days to process the depth, then fails on geometry build... great so I really can't do much except switch to High setting, or lower but switching to another level  photoscan has to re-build the depth data... would be great if you could preserve the "ultra high" depth data since it's there, and now worthless...

few other things that would be nice, is full picture display of depth, right now it's on the thumbnails only.. and even that there is depth data generated when you right click and want to export depth, photoscan once again has to rebuild the depth..... shouldn't it just export what took 2 days to generate in the first place?

3

General / Re: Cache directory in Photoscan

« on: April 10, 2012, 08:36:11 PM »

not sure if this helps any, I just finished building a model that was about 50mill poly. wanted to get as much
detail as possible, input was about 45 images from a 5D mark 2, 21 megapixel, using ultra high reconstruction, I was getting instant crashing on my machine with 48gig ram, luckily the mother board supported up to 192gig ram, so managed to fill it up to capacity, final usage was around 130 gig, there is definatly an issue with virtual memory, as soon as it hit's swap it almost always crashes on me.

4

Python and Java API / issue with Photoscan.Model

« on: March 21, 2012, 10:00:51 PM »

not sure if this is user error or a bug, but if I bring in a model using the "Import Model..." menu vs the python command, the model doesn't import into the proper location in space..

here's the basic way I am bringing in a model..

Code: [Select]

mdl = PhotoScan.Model()
mdl.load("c:/somemodel.obj",'obj')

chnk = doc.activeChunk
chnk.model = mdl

5

Python and Java API / Command line scripting support removed

« on: March 21, 2012, 09:52:47 PM »

why!!!?

6

Feature Requests / PTEX support

« on: March 07, 2012, 04:48:31 AM »

be nice not to have to deal with UV's... especially since it's almost impossible to re-do uv maps on any super high res geometry...

http://ptex.us/

7

General / Re: Using Chunks workflow...

« on: March 03, 2012, 02:33:10 AM »

Thanks for the tip! yes that does work, allowing me to merge the model..
I did end up just optimizing the amount of images I used for the reconstruction, using more than 50 images I would run out of memory during geometry build..

just some interesting specs..

system
dual hex-core xeon
96gb ram
3x quadro 4000

processed 37 full res images from a canon 5D mk2
ultra high reconstruction
processing time 16hrs
output geometry, 30million polys
texture map size, 20k x 20k

idea for future version, perhaps the ability to breakup a reconstruction volume into multiple builds (tiles)
to make it easier to digest for ultra high detail work.....

8

General / Using Chunks workflow...

« on: March 01, 2012, 02:39:18 AM »

here's a observation based on my last photoscan test...

1. I have a scene that I am reconstructing using about 60 images, I need to do "ultra" quality.

2. if I put all images into one chunk, alignment is perfect but "ultra" quality fails due to lack of ram, etc..

3. I break scene up into 4-5  image chunks, each chunk reconstructs fine, I then use Align chunks...
alignment completely screws up

4. put all images into once chunk, align photos, duplicate chunk 12 times, remove non used photos per chunk, now build geometry, everything is fine, but now can't merge the chunks because photoscan wants me to align them, even though they are already aligned! so just for the hell of it, I try to run align chunks, and once again about 1/2 the chunks are scrambled... I would like to be able to merge chunks even though it thinks they are not aligned.

also why does photoscan detect points on each image again when doing chunk alignment shouldn't those points been saved in the initial phase when building each chunk? seems a little redundant.

seems like an easy solution would be to allow you to merge chunks without requiring the alignment.... granted you follow the above workflow 

9

Feature Requests / Re: Split depth generation as separate task?

« on: February 28, 2012, 08:53:55 PM »

or even just have photoscan SAVE after generating depth and before generating the mesh... just had a photoscan running for over 86 hours that failed on mesh generation.... now it's back to square one 

10

Feature Requests / Split depth generation as separate task?

« on: February 16, 2012, 11:30:38 PM »

would it make sense to have the cpu/gpu intensive task of generating the depth data in each image
as a separate process from the actual mesh generation? say you make a target mesh of 100000 polys
but then want to have a higher res mesh, could just re-use the saved depth data go straight to the
geometry construction phase?...   just spent 24hrs doing depth generation, wish it could be re-used..

11

Feature Requests / Rotate Coordinate System.

« on: January 05, 2012, 10:32:41 PM »

handy feature, lots of times when I build geometry and export an .obj the model is upside down or in some strange orientation, a simple tool that would allow you do visually orient the geometry in relation to x,y,z axis would be great, with perhaps a ground plane grid display.. another option would be to be able to pick say 3 points to define an axis plane

12

Feature Requests / isolated points cleaning feature..

« on: November 02, 2011, 04:16:28 AM »

sometimes when reconstructing a dataset you get a lot of noisy points just floating in random locations, usually these points are pretty isolated to themselves, a neat feature would to be able to define a isolation kill radius (after allignment), changing this radius it would search from each point to the next nearest point, and if it doesn't have any points within that radius, or even a thrshold like if a point has less than X neighbor points it gets removed this should help create a cleaner geometry build..

13

Feature Requests / Flag images for texturing.

« on: July 22, 2011, 02:35:39 AM »

be a good feature to be able to flag certain images to be or not be used in the texture map generation.
example: I'm using 6 cameras for facial performance capture, 4 of the cameras are b/w cams, used only
for geometry creation, but I would like to flag the other 2 color cameras to just be used for texture creation.

14

Feature Requests / Re: resume batch.

« on: July 07, 2011, 08:42:06 AM »

yes for the moment I only have 2 images per chunk, as the cameras are in a fixed configuration, the cameras have sync'd shutters to capture at exactly the same time.

 I will expand the system shortly, did some testing, seems about 6 cameras should give me pretty decent coverage of about 180 deg of a persons face, I also have the capability to capture at over 100 frames per second, as to be able to temporally average out noise in the capture data, since stereo reconstruction tends to be noisy for this type of application.

 I intend to do a write up and post some examples once I am further in my testing.

15

Feature Requests / Re: resume batch.

« on: July 06, 2011, 07:14:23 PM »

here's my process.
I am currently testing using photoscan to do facial capture, I have builta small test rig currently using only 2 cameras (eventually 10 or more).

I bring in each pair of images into a new chunk of the sequence that I am processing, I then import cameras from a previous exported frame, so each alignment happens in the same world space.

I've only been testing with about 100 chunks at at time currently, since it's a bit labor intensive. altho I did create a helper utiltiy that aid's in setting up the initial .psz scene. (just loading images into chunks)

Alignment is not being run in batch since I have to import cameras for every chunk. crash happens when building geometry, it happens in batch and if I manually build, somewhere between 20 or 30 chunks further once I restart, it seems to crash. batch texture generation appears to work properly. 

I am running windows 7, on a i7-2600k cpu, w/Nvidia GTX580 driver ver 275.33

:UPDATE:
I just ran a test disabling OpenCL in preferences and it appears that the crashes do not occur, it did take quite
a lot longer to process but it did successfully finish.

I will send you my scene and image files that are causing the crash.


thanks and it is an truly awesome program....