Messages

31

General / Re: Best camera for Photoscan?

« on: January 14, 2015, 02:49:39 AM »

Technically it can use anything from goPro, iphone, compact, full dslr

It really depends on your purpose.

A top quality DSLR will give best results, but if your computer/gpu isnt strong enough to process the large images then you dont benefit. So in this situation if you want the best you can get its Camera+system sombination.

For a lower level than a top range camera, Im using the new sony rx100M3. Its small, light, makes a good picture, and easy to put on extension poles or copters. I can recommend it UNLESS you want to use photoscan pro and use GPS info, as this camera doesnt store GPS data

I have a canon D1100 also which is a fairly entry level dslr, but takes a nice photo also..and it came with 2 useful lens if you need to do some zoom photography (I do environments, so the ability to zoom a bit is useful)
Together with the rx100M3 its a nice combination of detail, clarity and portability.

I have 2 9xx series GPU cards and 32GB ram to crunch the info. My last photo recon of a cliff had a result of 1700 photo. Im aligning now (takes over night each attempt)
I also threw in a Canon s100 as that one has GPS and can hopefully lock the whole set to a world space since the other 2 cameras dont have GPS

-P

32

General / Re: CPU and GPU benchmarks

« on: November 16, 2014, 09:54:06 AM »

Have just built a new Rig for home

Intel i7-5930K CPU @ 3.50
Asus Ramapage V motherboard
32  GB Ram (for now)
Gigabyte 980GTX G1 4GD 
Gigabyte 970GTX G1 (Setup as eGPU inside a viDock -> expressCard PCie interface)

Wasnt sure what scene to benchmark so did the statue one. Ultra High and Aggressive.

Using device: GeForce GTX 970, 13 compute units, 4096 MB global memory
Using device: GeForce GTX 980, 16 compute units, 4096 MB global memory


finished depth reconstruction in 2288.13 seconds
Device 1 performance: 616.547 million samples/sec (GeForce GTX 970)
Device 2 performance: 1012.03 million samples/sec (GeForce GTX 980)

Using 1.1.0 pre-release (as current release didnt work with the 9xx series or nvidia drivers)
Hope that is useful info

Cheers

-Paul

33

General / Re: GPU processing during the modeling and settings

« on: November 16, 2014, 08:53:17 AM »

Hello pjenness,

Could you please try installing PhotoScan 1.1.0 pre-release (http://www.agisoft.com/forum/index.php?topic=2883.0) and run a short test to check if the problem is solved?

Success thankyou!!

And this is with a 980GTX installed, with a 970GTX in a viDock eGPU solution as a secondary GPU card. (as I share it between desktop and mobile macBook)

Running test scene to benchmark now

Cheers!!

-P

34

Bug Reports / GPU processing failed, switching to CPU mode

« on: November 16, 2014, 05:05:54 AM »

Hiya
I have a brand new Machine and install.

THe dense cloud process is failing on GPU

Code: [Select]

Using device: GeForce GTX 970, 13 compute units, 4096 MB global memory
  max work group size 1024
  max work item sizes [1024, 1024, 64]
  max mem alloc size 1024 MB
Using device: GeForce GTX 980, 16 compute units, 4096 MB global memory
  max work group size 1024
  max work item sizes [1024, 1024, 64]
  max mem alloc size 1024 MB
initializing...
selected 32 cameras from 32 in 0.129 sec
Loading photos...
Reconstructing depth...
[GPU] estimating 2976x4740x1728 disparity using 992x948x8u tiles, offset -454
[GPU] estimating 3233x4767x544 disparity using 1078x954x8u tiles, offset -145
ocl_engine.cpp line 320: clWaitForEvents failed, error -9999
GPU processing failed, switching to CPU mode
[CPU] estimating 3233x4767x544 disparity using 1078x954x8u tiles, offset -145
ocl_engine.cpp line 320: clWaitForEvents failed, error -9999
GPU processing failed, switching to CPU mode
Nvidia drivers are latest as of nov11

Any ideas?

Hope to get it going as this machine was built with photoscan in mind!

Cheers

-P

35

General / Re: GPU processing during the modeling and settings

« on: November 16, 2014, 04:57:24 AM »

HIya

Are people still getting GPU issues?

I have new win 7 build with 980GTX latest drivers (nov11) and getting the gpu failed issue

Cheers

-P

36

General / Re: CPU and GPU benchmarks

« on: May 11, 2014, 11:56:04 PM »

Hi @pjenness,
It would be interesting if you would also publish the associated GPU performances, ie samples per second, please.
Cheers / Tom

Hiya

I have update my post above to inlcude the samples.
Let me know if more infor needed

Cheers

-P

37

General / Re: If money doesn't matter what graphic card would you recommend for PhotoScan?

« on: May 09, 2014, 01:18:59 AM »

I benchmarked a few tha may be interesting for you

http://www.agisoft.ru/forum/index.php?topic=1868.msg12397#msg12397


if going nvidia I recommend GTX over Quadro.
The 780 GTX is really good value for speed at the moment:


-P

38

General / Re: GPU recommendation Quadro K2000D or gaming card?

« on: May 05, 2014, 12:04:15 AM »

can post some bench results on our dataset ? that our community can have dreams on new GPUs ?

Done.
Sorry took so long, I had to get NDA approval first , since it was internal testing.
Posted to the main benchmark thread.

-P

39

General / Re: CPU and GPU benchmarks

« on: May 04, 2014, 11:59:44 PM »

Hi.

Just sharing this if useful:

All these tests were run on

    Dell Precision T7600
    Intel® Xeon(R) CPU E5-2620 0 @ 2.00GHz ? 12 core
    Memory: 125.8 GB

GPU was turned on, 1 GPU to replace 1 CPU core as recommended.
Where more GPU than CPU available no CPU used.


Running the bench mark scene BUild Dense Cloud at Ultra High, Moderate Filtering


1x GeForce GTX 680 (8 Cores @ 1124Mhz, 2047MB)   Finished processing in 1289.69 sec (21min 29sec)
Device 2 performance: 632.531 million samples/sec (GeForce GTX 680)

1x GeForce GTX 780 (12 Cores @ 1019Mhz, 3071MB)  Finished processing in 984.936 sec (16m 25sec)
Device 1 performance: 823.705 million samples/sec ( GeForce GTX 780 )

1x GeForce GTX TITAN (14 Cores @ 928Mhz, 6143MB)     Finished processing in 959.494 sec (16m)
Device 2 performance: 840.561 million samples/sec (GeForce GTX TITAN)

2x Quadro 5000 (11 Cores @ 1026Mhz, 2559MB      Finished processing in 946.803 sec (15min 46sec)
Device 1 performance: 392.461 million samples/sec (Quadro 5000)
Device 2 performance: 396.724 million samples/sec (Quadro 5000)

1xGeForce GTX 780 (12 Cores @ 1032Mhz, 3071MB)
1xQuadro 5000 (11 Cores @ 1026Mhz, 2559MB                 Finished processing in 654.824 sec (10min 54sec)
Device 1 performance: 913.684 million samples/sec (GeForce GTX 780)
Device 2 performance: 397.382 million samples/sec (Quadro 5000)



Hope this is useful

Cheers

-P

40

Feature Requests / Re: Using PhotoScan as a Matchmoving Software

« on: May 04, 2014, 11:57:51 PM »

I use fbx to export hem all to maya.

Then made a simple script that select cameras, and run tool it makes 1 master camera fromm all the other cameras.
Just copy and key trans, rot, lens, filmback info etc.


As long as you keep everything in order and number the frames 1 - lastframe  then you get a nice animated camera.

Note you may have to put a -90 rotation in Y on the "Rotate Axis"  (This is not the same as the normal rotate channel, you will need to open the attribute editor in maya)

Cheers

-P

41

General / Output height map and color

« on: May 04, 2014, 11:53:41 PM »

Hi

Im wondering if Ive missed how to do this in the docs, or it just isnt a feature.


I want to use photoscan to make some texture tiles of ground at different distances. For example close up for pebble and stick details, mid for slight wide details and then maybe later wide, for aerial details.

Is there a way for photoscan to assume it is on a plane and generate height and colour that match?
So instead of going thru th mesh process the out put is height map and colour map from a top (or defined ) projection?

I notice one of the mesh options is height field, which meshs it based on an axis. But instead it would be kool to just generate a map of height and colour together.

Is this feature in there aready?

I imagine the process would be something like,

• Align photos of ground, or awll or whatever flattish object
• Place or define a plane thru the points at whcih will be the base height
• Run photoscan to generate the height map and colour map

Cheers

-P

42

Feature Requests / Re: Support for multiple UV tiles (nd Udim)

« on: April 14, 2014, 08:24:36 AM »

Hi

Just bumping this

Is this possible to add?

Just got a aset that has over 20 tiles of UVs , which means I have to manually extract texture 20 + times and shift the UVs to 1001 each time.

Cheers

-P

43

General / Re: GPU recommendation Quadro K2000D or gaming card?

« on: April 10, 2014, 08:57:05 AM »

Hi

Recently Ive benchmarked

Quadro 5000 (11 Cores @ 1026Mhz, 2559MB

2x Quadro 5000 (11 Cores @ 1026Mhz, 2559MB

GeForce GTX 680 (8 Cores @ 1124Mhz, 2047MB)

GeForce GTX 780 (12 Cores @ 1019Mhz, 3071MB)

GeForce GTX TITAN (14 Cores @ 928Mhz, 6143MB)


The 780GTX was WAY bang for buck!!  It  was seconds slower than the 2xK5000, or 1x Titan GTX

Right now Im sitting on 1x780GTX & 1 K5000 ...  with 128GB Ram. YAY!!   
Smashes heavy jobs.


-P

44

Feature Requests / Support for multiple UV tiles (nd Udim)

« on: April 10, 2014, 08:46:26 AM »

Hi

In vfx industry we are often using multiple UV tiles on an asset/object. This is to allow multiple 4k,8k maps of super resolution.

Could this be supported in photoscan please.?

At the moment I am currently trying to bake textures to a high res rock formation model that has 9 UV tiles. At the moment I have to reimport 9 times into PS with Uvs that shift thru the first default tile.

It would be great to adopt the conventions that mudbox and other apps do..

Give a starting tile (if the first tile is 1, or 1001 in our case) then counts up.
Also set a uDIm , the u  row max . For example a uDim of 10 will allow a row of 10 tiles, then stack to next row. (ie first row 1001-1010, the second row  1011-1020 ...)

If this could be native to PS that would be amazing. If not could settle for a python script for Pro version.

Thanks!

-P

45

Python and Java API / Camera isAligned, isCalibrated?

« on: March 02, 2014, 11:53:28 PM »

Hiya

Im looping thru cameras in my active chunk and want to test each one if they are aligned (maybe later if calibrated)

is there an attribute or function that I can acesss?

Something like:

Code: [Select]

import PhotoScan

doc = PhotoScan.app.document
myActiveChunk = doc.activeChunk
myCameras = myActiveChunk.cameras

for eachCamera in myCameras:
   
   if eachCamera.isAligned:
         
          print ('I am aligned ' )

Looking thru the api I cant find anything. Only an int for how many cameras aligned in a chunk.