Topics

1

General / AMD 5700XT support ( no SPIR support)

« on: July 17, 2019, 05:07:13 PM »

When I run my project, I see this:

AMD Radeon RX 5700 XT (gfx1010): no SPIR support

Will it affect the performance or results of the project?

2

Bug Reports / cudaGetDeviceCount failed: unknown error (30)

« on: July 15, 2019, 03:52:34 PM »

I am using Windows 10 with Radeon VII GPU
When I do Align Photos, I see a strange message:cudaGetDeviceCount failed: unknown error (30)
It is an AMD card, how come it will generate a cuda error? There is no nvidia card in the computer.

I have used nvidia cards on this computer with nvidia drivers installed, but Agisoft still should not pop this error as I have no cards in the computer now.

3

General / Bad Allocation, Should I continue?

« on: July 03, 2019, 10:09:10 AM »

I have a project running network mode.
The server has 256GB ram and it is fully used now 256GB/256GB
Project status is 88% in building Dense Cloud at this moment

It just fired a "Bad Allocation" message but the node and server seem can continue to progress.

Will there be any damages to the dense cloud if I continue the process?
SHould I terminate or let it continue? It still has 10 hours to wait if I continue, if I know it will fail finally I will stop processing now

Thank you

4

Feature Requests / Estimate Photo Quality Multi-node Network Processing

« on: June 27, 2019, 11:43:38 PM »

Now Estimate Image Quality function do not spread tasks to multiple nodes, it just runs on the one node, which is not efficient.

5

Bug Reports / unknown error (30) at line 33

« on: June 09, 2019, 12:35:16 PM »

I used Network mode to build Dense Cloud, I read this message in Network Monitor:
[17:01:20] unknown error (30) at line 33

What is this? This error is not attached to a node, but attached to the project in the upper pane

6

Bug Reports / Can't open file: The system cannot find the file specified (Cannot read Points)

« on: October 01, 2018, 01:45:12 PM »

1. I used network mode to process project. The Project has 1000 images with GPS tag. Using latest V1.4.4. 6848
2. The first alignment worked OK, both worker machines can read images from the network drive and process accordingly.
3. But I am not satisfied with the results.
4. Then I do this:
i. RIght click on  "Tie Points" at the Workspace pane.
ii. Click "Delete Tie Points"
5. Restart the Align Photos process by clicking Align Photos.


Then I see this:

2018-10-01 18:29:57 [192.168.1.124:54740] finished #9 MatchPhotos
2018-10-01 18:29:59 [192.168.1.124:54740] finished #9 MatchPhotos.initialize (1/1)
2018-10-01 18:30:00 [192.168.1.145:63411] failed #9 MatchPhotos.detect (2/62): Can't open file: The system cannot find the file specified (2): //192.168.1.177/fastraid/3rdTryTIF.files/0/0/point_cloud/points.grp
2018-10-01 18:30:01 [192.168.1.145:63411] failed #9 MatchPhotos.detect (3/62): Can't open file: The system cannot find the file specified (2): //192.168.1.177/fastraid/3rdTryTIF.files/0/0/point_cloud/points.grp
2018-10-01 18:30:02 [192.168.1.145:63411] failed #9 MatchPhotos.detect (4/62): Can't open file: The system cannot find the file specified (2): //192.168.1.177/fastraid/3rdTryTIF.files/0/0/point_cloud/points.grp
2018-10-01 18:30:02 [192.168.1.145:63411] failed #9 MatchPhotos.detect (5/62): Can't open file: The system cannot find the file specified (2): //192.168.1.177/fastraid/3rdTryTIF.files/0/0/point_cloud/points.grp
2018-10-01 18:30:22 [192.168.1.145:63411] finished #9 MatchPhotos.detect (6/62)
2018-10-01 18:30:27 [192.168.1.124:54740] finished #9 MatchPhotos.detect (1/62)

If I continue the process till the end, it will yield a Sparse Cloud with 9 tie points only.

If I also delete the Key Points, results are exactly the same.

The only way I can use Network mode again to get through Align Photo is to recreate a new project.

I can use standalone mode to do Align Photo if this error pop out, but it will still yield a Sparse Cloud with 9 tie points only.

So there will be no way to recover this project after this error.
 

7

Bug Reports / #0: CUDA_ERROR_UNKNOWN_CODE_719 (719) at line 123

« on: July 29, 2018, 05:54:03 PM »

I am using 1.4.3 Build6529
I got this error in a network processed job.
I have 2 computers in this setup.

accuracy = Highest, preselection = none, keypoint limit = 50000, tiepoint limit = 7000, apply masks = 0, filter tie points = 0

Job cannot pass through alignment stage and got stuck at 98% Matching Photos

8

General / AMD Vega64 vs Nvidia GTX 1080 vs GTX 1080Ti

« on: October 10, 2017, 12:55:19 PM »

Can anyone or Agisoft tell, which card is faster when running Photoscan?

Assume all other factors constant.
We are building a new machine and speed is the key, don't care about the cost first.

9

Bug Reports / PS1.3 Build 3520 Error: out of memory at line 177

« on: March 05, 2017, 02:34:27 PM »

I have installed latest driver of R9 390X and GTX780
When I run Align Photos with a 500 images dataset, I get this error.
I can process if I uncheck all GPU options.
I have tried reinstalling PS1.3 , nvidia and amd drivers, problem persists.


2017-03-05 19:26:53 Agisoft PhotoScan Professional Version: 1.3.0 build 3520 (64 bit)
2017-03-05 19:26:53 Platform: Windows
2017-03-05 19:26:54 OpenGL Vendor: ATI Technologies Inc.
2017-03-05 19:26:54 OpenGL Renderer: AMD Radeon (TM) R9 390 Series
2017-03-05 19:26:54 OpenGL Version: 4.5.13469 Compatibility Profile Context 21.19.519.2
2017-03-05 19:26:54 Maximum Texture Size: 16384
2017-03-05 19:26:54 Quad Buffered Stereo: not enabled
2017-03-05 19:26:54 ARB_vertex_buffer_object: supported
2017-03-05 19:26:54 ARB_texture_non_power_of_two: supported
2017-03-05 19:27:00 LoadProject
2017-03-05 19:27:00 Loading project...
2017-03-05 19:27:13 loaded project in 12.674 sec
2017-03-05 19:27:13 Finished processing in 12.675 sec (exit code 1)
2017-03-05 19:27:36 AlignPhotos: accuracy = Highest, preselection = generic, reference, keypoint limit = 40000, tiepoint limit = 4000, adaptive fitting = 1
2017-03-05 19:27:36 Detecting points...
2017-03-05 19:27:36 Using device: GeForce GTX 780, 12 compute units, 3072 MB global memory, CUDA 3.5
2017-03-05 19:27:36   max work group size 1024
2017-03-05 19:27:36   max work item sizes [1024, 1024, 64]
2017-03-05 19:27:36 Using device: Hawaii, 44 compute units, 8192 MB global memory, OpenCL 2.0
2017-03-05 19:27:36   max work group size 256
2017-03-05 19:27:36   max work item sizes [256, 256, 256]
2017-03-05 19:27:36   max mem alloc size 4048 MB
2017-03-05 19:27:36   wavefront width 64
2017-03-05 19:27:40 Device 1
2017-03-05 19:27:40    Program build log:
2017-03-05 19:27:40 Error: Undeclared function index 1025
2017-03-05 19:27:40
2017-03-05 19:27:40
2017-03-05 19:27:40 Finished processing in 4.368 sec (exit code 0)
2017-03-05 19:27:40 Error: out of memory at line 177
2017-03-05 19:28:16 DuplicateChunk: items = ()
2017-03-05 19:28:16 Duplicating chunk...
2017-03-05 19:28:18 Finished processing in 1.2 sec (exit code 1)
2017-03-05 19:29:24 AlignPhotos: accuracy = Highest, preselection = generic, reference, keypoint limit = 40000, tiepoint limit = 4000, adaptive fitting = 1
2017-03-05 19:29:24 Detecting points...
2017-03-05 19:29:24 Using device: GeForce GTX 780, 12 compute units, 3072 MB global memory, CUDA 3.5
2017-03-05 19:29:24   max work group size 1024
2017-03-05 19:29:24   max work item sizes [1024, 1024, 64]
2017-03-05 19:29:24 Using device: Hawaii, 44 compute units, 8192 MB global memory, OpenCL 2.0
2017-03-05 19:29:24   max work group size 256
2017-03-05 19:29:24   max work item sizes [256, 256, 256]
2017-03-05 19:29:24   max mem alloc size 4048 MB
2017-03-05 19:29:24   wavefront width 64
2017-03-05 19:29:28 Device 1
2017-03-05 19:29:28    Program build log:
2017-03-05 19:29:28 Error: Undeclared function index 1025
2017-03-05 19:29:28
2017-03-05 19:29:28
2017-03-05 19:29:28 Finished processing in 3.829 sec (exit code 0)
2017-03-05 19:29:28 Error: out of memory at line 177

10

General / Should we use Lens Profile Correction in Photoshop before importing into PScan?

« on: December 29, 2015, 03:54:06 PM »

I know PScan can undistort photos automatically using the detected feature points during the Align Photos step.
But what I am confuse is that if our camera has a built in lens profile in Adobe Lightroom, like Sony DSC-RX1, when we export the photos after exposure adjustments, should we also apply the lens correction profile to undistort the photos before importing the JPGs into PScan?
Will this affect the accuracy of the output of DEM and DOM? Better or Worse?

11

Bug Reports / Bad Allocation in Building Texture and a series of trailing problems

« on: July 12, 2014, 10:06:09 AM »

I have a problem when using Photoscan Professional 1.0.4 build 1847
I am processing around 187 images with 5616*3744 resolution

From workspace I see this:
182cameras, 7 markers, 2797815points
Dense Cloud 142591765 points
3D Model 28370252 faces
Vertices 14192523

When I start to build Texture I selected:
Building Texture
Orthophoto
Mosaic
Texture size:
90000 x 1
Enable Colour Correction

I have a bounding box very tight
My computer has 32GB Ram, i5-2400 CPU and a Radeon 7850 with 2G Ram

What I got is Bad Allocation Error in the console.
The software did not prompt a dialog box but it just go into a "Not Responding" mode
While CPU usage is held around 60% and Ram usage is around 5G when it goes into "Not Responding" mode, also OpenCL shows GPU usage is around 60% too.

So I force closed the program. Now when I open the project again, the program becomes very slow and sometimes windows display "Program Not Responding" and it will recover after some random time and go into "Program Not Responding" again. The model view cannot be rotated anyway and when I drag the mouse, the program goes into  "Program Not Responding" and need to wait for a long time to recover.

The program consumes 60% of GPU and 70% of CPU and 6G Ram as long as the project loads and when I try to start building texture again, it will stay in Progress 0% forever.

12

Camera Calibration / Minimum size of calibration pattern and near limit of camera

« on: July 05, 2014, 11:49:50 AM »

Hi,
We need to calibrate a Full Frame DSLR Camera using 35mm lens for aerial survey.
We will fly at a height of 300m so we set the focus to a point that the subject distance is 300m and the near limit of Depth of field is around 5m.

That means within a 5m distance, our camera will not be able to acquire a sharp image.
I see that the lens calibration software is using the monitor to display the checker board, since the monitor is small, if I use this lens focus ring setting to calibrate, I will get a blurry result if I do not turn the ring to refocus on the screen.

I want to know, if the calibration still works if I calibrate with the same focus ring setting for a 300m subject and acquire a blurry image of the checker board or should I turn the focus ring to make a sharp image when doing the calibration?