Agisoft PhotoScan 1.1.0 pre-release

[KoStyaN]

Is it possible save fisheye as default calibration?

[anaana]

Hello,
I have some problems with panorama view, when I create a new camera group put my photos there and aligning them i i see nothing at mt model display.... Did somebod already try this option?

[Alexey Pasumansky]

Hello anaana,

And have you switched camera group type to the Camera Station in the Workspace pane prior to alignment process?

[anaana]

Yes, i created the camera group and than ser the group type to Station, after i clicked to align photos and at the end of process i had nothing on me model display.

[Alexey Pasumansky]

Hello anaana,

Then maybe you can provide us this problematic dataset on support@agisoft.com, so that we can check the problem?

[anaana]

Ok, thank you.

[eriks]

Photoscan 1.1.0 build 1985 now gives the error shown below and a bluescreen "System_Service_Exception" when building a pointcloud out of 7 chunks.
a brand new PC(intel i7-5930) with windows 8.1 and 2x GeForce GTX 980 (NVIDIA driver 344.48)
have 64Gb ram, 10 of 12 cpu cores active, and the two graphics cards active as OpenCL.
is it software or hardware failure?


1 blocks: 170
calculating match counts... done in 0 sec
overlapping groups selected in 0 sec
1 blocks: 170
final block size: 170
adding 10951 points, 144 far (12.272 threshold), 0 inaccurate, 15 invisible, 15 weak
3 sigma filtering...
adjusting: xxxxxxxx 1.79733 -> 1.79721
point variance: 1.79934
adding 14 points, 779 far (5.39803 threshold), 0 inaccurate, 7 invisible, 5 weak
adjusting: xxxxxxxxxxx 1.26301 -> 1.24688
point variance: 1.24836
adding 107 points, 689 far (3.74508 threshold), 0 inaccurate, 6 invisible, 6 weak
adjusting: xxxxxxxxxxx 1.07205 -> 1.06824
point variance: 1.06955
adding 387 points, 726 far (3.20864 threshold), 1 inaccurate, 6 invisible, 5 weak
adjusting: xxxxxxxxxxx 0.996641 -> 0.995105
point variance: 0.996358
adding 595 points, 757 far (2.98908 threshold), 1 inaccurate, 5 invisible, 3 weak
adjusting: xxxxxxxx 0.965478 -> 0.964742
point variance: 0.965969
adding 709 points, 795 far (2.89791 threshold), 1 inaccurate, 5 invisible, 5 weak
optimized in 1.954 seconds
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxxxxx 0.883686 -> 0.844905
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxxxxx 0.910176 -> 0.89795
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxxx 0.929815 -> 0.925034
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxxxx- 0.94085 -> 0.929951
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxxxxxxxxx 0.938584 -> 0.932644
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxx 0.937024 -> 0.931951
coordinates applied in 0 sec
Optimizing point cloud...
adjusting: xxxxxxxxxxxxxx 0.947859 -> 0.941533
coordinates applied in 0 sec
Saving project...
saved project in 1.594 sec
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
clGetDeviceInfo(CL_DEVICE_GLOBAL_MEM_SIZE) failed: CL_OUT_OF_RESOURCES
ocl_engine.cpp line 177: clGetDeviceInfo(device_id, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(max_mem_alloc_size), &max_mem_alloc_size, NULL) failed, error CL_OUT_OF_RESOURCES
Saving project...
saved project in 1.531 sec
Finished processing in 44855.2 sec (exit code 1)
Agisoft PhotoScan Version: 1.1.0 build 1985 (64 bit)
OpenGL Vendor: NVIDIA Corporation
OpenGL Renderer: GeForce GTX 980/PCIe/SSE2
OpenGL Version: 4.4.0 NVIDIA 344.48
Maximum Texture Size: 16384
Quad Buffered Stereo: not enabled
ARB_vertex_buffer_object: supported
ARB_texture_non_power_of_two: supported

[r0xx]

Is there any best practice to automatically close a node without causing errors? Is there any command to stop it, let it finish and then close without me having to force kill it?

[Alexey Pasumansky]

Hello eriks,

Could you please run any stress tests for OpenCL computations? Unfortunately, we still haven't yet got test configuration with GTX980, so can't check if the issue is reproducible.

[ppkong]

It's very useful that add export tiles function in photoscan 1.1.But I think there are several points to continuous improvement.
1.Python Scripting should be support output tiles;
2.Output tiles in ps1.1 version speed significantly slower than the ps1.0;
3.After output tiles must be use Agisoft Viewer to open it and convert to other formats,and every time you can convert only one *.TLS file,and the number of tiles mesh cannot be based on actual model number of the output;
So I think it should support output data different format;

[eriks]

hi, yes, I have been running some stress-tests on the dual GTX980's getting their core temperature quite high above 80degrees with still no problems. But that test was   https://unigine.com/products/heaven/   which I think tests DirectX.
Do you have an recommended OpenCL stress test, or is    http://www.geeks3d.com/20140917/msi-kombustor-3-5-0-gpu-burn-in-stress-test-opengl-opencl-benchmark  possibly an good test? 
At the moment, There are some 55 hours remaining on another process on that computer so I am not touching it until Monday.

thanks, ErikS

[Alexey Pasumansky]

Hello Erik,

Since the problem is related to OpenCL it's better to run OpenCL test. We can't recommend any special test, so probably the one you've mentioned would be fine.

In the related topic you can find that GTX 980 works fine on the latest nVidia drivers with PhotoScan 1.1.0: http://www.agisoft.com/forum/index.php?topic=1868.msg16292#msg16292

[StevenF]

PhotoScan.Chunk.copy() is causing PhotoScan to crash on v 1.1.0 build 1985 (64 bit). I submitted a crash report.
Here is a simple work around in case anyone else working with the beta is having this problem:

doc = PhotoScan.app.doc

# create a copy of the document
doc2 = PhotoScan.Document()
doc2.open(doc.path)

# append to current document
doc.append(doc2)

# remove unwanted chunks if necessary
doc.remove(doc.chunks[-1])

[Alexey Pasumansky]

Hello Steven,

Thanks for reporting, however, we haven't been able to reproduce the problem with crashes.

But as for the creating duplicated copies of chunks and adding them to the existing document, we will try to implement this functionality in Python API.

[James]

I can't see any perceivable difference when either texturing or generating orthophoto, with or without colour correction enabled, in Photoscan Pro 64 bit build 1985 on windows 8.1.

At first I thought it was working because the colours seemed nice and even although the photos were fairly poor, but there were still some seams visible. So i tried with colour correction disabled and I couldn't see any difference. I tried both orthophoto and texturing with mosaic blending (4096x4). The processing time seems very similar too, which surpised me as colour correction used to take ages.

I am working with ~200 ~9MP TIF images.