Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Topics - MeHoo

Pages: 1 [2]
Bug Reports / Nevermind.. delete me
« on: May 06, 2015, 02:56:35 AM »
EDIT: Corrupt server drive

Feature Requests / please work with these guys...
« on: March 31, 2015, 11:45:59 PM »
Euclideon Geoverse 2013:

Long story short.. I shot video and took pictures.  They both solve separately, but what is the best method for merging them together?  Dense clouds?  Mesh?

Just curious if others have gotten better results either way.  Thanks.

Feature Requests / Batch Process - Add "Load file" or...
« on: January 12, 2015, 12:06:44 AM »
Ok so here's my current dillemma.  I'm doing a bunch of testing using batch to process things differently and save off chunks to versioned files so I don't destroy data.  I'd love for photoscan to keep the opened file current.  For instance, I've told batch to save the project as v0009 and this is coming from v0007.  After it batches, it still says v0007 is open eventhough technically the data that is present is a result of v0009's options. 

I can see the logic here.  It's a great way to fire off revisions, but I'd love the option to load the test versions.  Basically I want to batch 5 different processing options which requires opening/saving versioned files.

Can this be added?  Thanks!

None of my GPUs show up in the openCL dialog.  Is this merely a display glitch or are they disabled somehow due to the remote desktop connection?

I built a monster of a machine, 256gigs of ram, dual quadcores, 2x 580 GTXs, 1x Quadro K4000... but it's kind of a server, and my best option is to remote in and operate, but I am concerned that the GPUs aren't working right.  What's the best way to check if this is just a display bug?  When I go to openCL there is nothing shown, but If I manually go to the box, it shows all three GPUs fine.

I'm only running the standard edition because I'm an artist, and the pro version is insanely expensive for what I'm doing, so perhaps this is already possible...

So I'm testing building a multiple gopro rig for on-location scans of buildings and objects.  I figure, if I can tell PScan exactly how these cameras are configured, wouldn't that increase accuracy of the scans and speed of the processing somehow?  I'd assume the alignment process would greatly improve if it knew their relationship as a constant.

Also, being able to load photosets from both cameras as "pairs" based on their folder structure would be very helpful, so image_0001 in each folder would be treated as a pair from this grouping.  MY GoPro's are wifi linked so they will be taking photos at exactly the same time.

Just a thought.  I'd like to hear if this would be helpful or not.

- Justin

Bug Reports / Is there a second bounding box I'm not aware of..?
« on: January 03, 2015, 02:57:02 AM »
So the aligned cameras on this long sweep of boulders that I did shows up fine, and the points seem to be accurate, but when I build the dense point cloud or mesh, they get clipped like my bounding box wasn't set large enough.  I've re-sized the box numerous times, but after a certain size it gets a clean line chopped off.  I did not notice it splitting any chunks, it just has no data after that imaginary wall.  Any ideas?  Something I'm doing wrong?  I've not encountered this before.

Standard edition, latest build.  I'll get you more info when I can, as the machine is currently processing a 100+ hour photo alignment (stress testing).

- Justin

Getting these long, disk-paging scans  to get to 100% after 24+ hours, but then they just stop at 100% for many hours. 

The Overall progress bar still has about a 15% gap at the end, and I can see my paged memory (Commit Size) dropping and moving, etc, but it doesn't seem like it will ever end.  Kind of annoyed that 100% isn't 100%.  Safe to assume that if the Commit Size and memory are still bouncing around that this isn't locked up?

Also, does Photoscan have an option to auto-save after each big step in the processing in case you need to kill the process?  Resuming this on a rented 200 gig machine would be awesome, for instance.

Any help is appreciated.  :/

I will say this, the latest build is working amazingly well for alignments.  Pretty happy with that, and the ability to build the dense cloud now.  I wish there was the ability to use distributed computing to build these meshes though, as I have servers sitting idle while one machine chomps away at it.

I have dual GTX 580's.  Both are checked in preferences.  Two cores on my 6-core i7 are disabled as instructed. 

No GPU load is detected whatsoever during either photo alignment or mesh generation.

I'd love to utilize these monsters to do some calculations... Any help is appreciated.  What am I doing wrong?

General / Help with camera setup?
« on: January 16, 2012, 07:50:31 AM »
I shoot with a Canon 7D and most of my images that I'm trying to process were shot with the 100mm f/2.8 L Macro IS lens.  For some reason when I load them into photoscan and align them, they stretch out in the "z" depth really badly.  Images shot the same day with the 5d MK2 do not do this.  Is there a calibration step that I am missing?  Is there some calibration file I can load for a 7d?

Do I not have enough coverage and parallax?

Also can someone explain to me real fast.. when masking images, do you want the UNSHADED area to be the area that gets processed, or the shaded?  I assumed the unshaded area is what gets processed.

Everytime I try masking, this problem gets worse.

Thanks in advance for any help.

Feature Requests / Just a few (hopefully) simple requests.
« on: January 14, 2012, 08:00:23 AM »
Loading 8bit jpegs is fine for generation, but having the simple option to swap to higher res TIFFs and keeping the 12bit+ depth for texture generation would help immensely.  This could be as simple and rudimentary as having an "alternate" folder with the exact same filenames with different extensions possibly.

I would also love to see a manual targeting mode for the standard edition.  I have photos that I know have overlapping coverage, but for whatever reason, photoscan isn't seeing it.  I'd love to be able to just give it a few markers, and say A and B on two images, and have it try again.  Most softwares like this have that option in some form or another.  At least for aligning chunks.

The last option that would be really helpful is to see what a selected images' contribution to the pointcloud/mesh is.  Picking an image and seeing that it is the offending image in the generation would make it much easier to delete the problem areas.

Love the software.  I hope that you can help us out with these things.  Thanks! :)

Pages: 1 [2]