Agisoft Metashape 1.7.0 pre-release

[Alexey Pasumansky]

Hello Shadow,

Version 1.7.0 has been published on Downloads page:
https://www.agisoft.com/downloads/installer/

Feel free to share your feedback on the latest version.

[flyzk]

Hi, happy new year ! When can we expect a upgrade of the pdf manual for 1.7 ?

[JMR]

Happy new Metashape version!
Exclude stationary points... sounds like a great help for rotary table scans, doesn't it?

[Alexey Pasumansky]

Hi, happy new year ! When can we expect a upgrade of the pdf manual for 1.7 ?

Hello flyzk,

The PDF manual will be published later in January. Meanwhile the manual prototype is available as built-in application help.

[lunar]

  Continuous shutdown of the application during building the mesh with depth maps - during calculating depth maps.
new nvidia drivers uploaded
final build 1,7

[Alexey Pasumansky]

Hello lunar,

Please enable Write Log to File option in the General Preferences tab, re-start Metashape and if the issue persists, send the saved log to support@agisoft.com.

[eltama]

Hello Alexy
Same error here, metashape quit without error in the log fle, mesh set  in highest and high quality:

2021-01-03 03:24:41 Agisoft Metashape Standard Version: 1.7.0 build 11736 (64 bit)
2021-01-03 03:24:41 Platform: Windows
2021-01-03 03:24:41 CPU: AMD Ryzen Threadripper 1950X 16-Core Processor (desktop)
2021-01-03 03:24:41 CPU family: 23 model: 1 signature: 800F11h
2021-01-03 03:24:41 RAM: 127.9 GB
2021-01-03 03:24:42 OpenGL Vendor: NVIDIA Corporation
2021-01-03 03:24:42 OpenGL Renderer: GeForce RTX 2080 SUPER/PCIe/SSE2
2021-01-03 03:24:42 OpenGL Version: 4.6.0 NVIDIA 460.89
2021-01-03 03:24:42 Maximum Texture Size: 32768
2021-01-03 03:24:42 Quad Buffered Stereo: not enabled
2021-01-03 03:24:42 ARB_vertex_buffer_object: supported
2021-01-03 03:24:42 ARB_texture_non_power_of_two: supported
2021-01-03 03:24:46 Loading project...
2021-01-03 03:24:49 loaded project in 3.165 sec
2021-01-03 03:24:49 Finished processing in 3.198 sec (exit code 1)
2021-01-03 03:25:06 checking for missing images...Checking for missing images...
2021-01-03 03:25:12  done in 5.853 sec
2021-01-03 03:25:12 Finished processing in 5.853 sec (exit code 1)
2021-01-03 03:25:12 BuildModel: quality = High, depth filtering = Mild, PM version, source data = Depth maps, surface type = Arbitrary, face count = High, volumetric masking = 0, OOC version, interpolation = Enabled, vertex colors = 1
2021-01-03 03:25:12 Generating depth maps...
2021-01-03 03:25:12 Preparing 9964 cameras info...
2021-01-03 03:25:26 cameras data loaded in 13.268 s
2021-01-03 03:26:12 cameras graph built in 46.628 s
2021-01-03 03:26:12 filtering neighbors with too low common points, threshold=50...
2021-01-03 03:26:12 Camera 3878 has no neighbors
2021-01-03 03:26:12 Camera 3888 has no neighbors
2021-01-03 03:26:12 Camera 4026 has no neighbors
2021-01-03 03:26:12 Camera 4571 has no neighbors
...
2021-01-03 03:26:12 avg neighbors before filtering: 634.779 (90% filtered out)
2021-01-03 03:26:12 limiting neighbors to 40 best...
2021-01-03 03:27:32 avg neighbors before limiting: 61.1616 (55% limited out)
2021-01-03 03:27:32 neighbors number min/1%/10%/median/90%/99%/max: 0, 0, 6, median=37, 40, 40, 40
2021-01-03 03:27:32 cameras info prepared in 139.811 s
2021-01-03 03:27:35 saved cameras info in 3.159
2021-01-03 03:27:35 Partitioning 9964 cameras...
2021-01-03 03:27:35 number of mini clusters: 767
2021-01-03 03:27:35 256 groups: avg_ref=38.9219 avg_neighb=187.617 total_io=582%
2021-01-03 03:27:35 max_ref=39 max_neighb=381 max_total=420
2021-01-03 03:27:35 cameras partitioned in 0.071 s
2021-01-03 03:27:35 saved depth map partition in 0.011 sec
2021-01-03 03:27:39 loaded cameras info in 3.503
2021-01-03 03:27:39 loaded depth map partition in 0.01 sec
2021-01-03 03:27:39 already partitioned (38<=50 ref cameras, 184<=200 neighb cameras)
2021-01-03 03:27:39 group 1/1: preparing 222 cameras images...
2021-01-03 03:27:39 point cloud loaded in 0.086 s
2021-01-03 03:27:40 Found 2 GPUs in 0.174 sec (CUDA: 0.063 sec, OpenCL: 0.111 sec)
2021-01-03 03:27:40 Using device: GeForce RTX 2080 SUPER, 48 compute units, free memory: 6646/8192 MB, compute capability 7.5
2021-01-03 03:27:40   driver/runtime CUDA: 11020/8000
2021-01-03 03:27:40   max work group size 1024
2021-01-03 03:27:40   max work item sizes [1024, 1024, 64]
2021-01-03 03:27:40 Using device: GeForce RTX 2080 SUPER, 48 compute units, free memory: 6693/8192 MB, compute capability 7.5
2021-01-03 03:27:40   driver/runtime CUDA: 11020/8000
2021-01-03 03:27:40   max work group size 1024
2021-01-03 03:27:40   max work item sizes [1024, 1024, 64]
2021-01-03 03:29:10 group 1/1: cameras images prepared in 90.406 s
2021-01-03 03:29:10 group 1/1: expected peak VRAM usage: 4970 MB (775 MB max alloc, 9630x14397 mipmap texture, 40 max neighbors)
2021-01-03 03:29:10 group 1/1: 222 x frame
2021-01-03 03:29:10 group 1/1: 222 x uint8
2021-01-03 03:29:10 Found 2 GPUs in 0 sec (CUDA: 0 sec, OpenCL: 0 sec)
2021-01-03 03:29:10 Using device: GeForce RTX 2080 SUPER, 48 compute units, free memory: 6639/8192 MB, compute capability 7.5
2021-01-03 03:29:10   driver/runtime CUDA: 11020/8000
2021-01-03 03:29:10   max work group size 1024
2021-01-03 03:29:10   max work item sizes [1024, 1024, 64]
2021-01-03 03:29:10 Using device: GeForce RTX 2080 SUPER, 48 compute units, free memory: 6693/8192 MB, compute capability 7.5
2021-01-03 03:29:10   driver/runtime CUDA: 11020/8000
2021-01-03 03:29:10   max work group size 1024
2021-01-03 03:29:10   max work item sizes [1024, 1024, 64]
2021-01-03 03:29:10 [GPU 1] group 1/1: estimating depth map for 1/38 camera 225 (9 neighbs)...
2021-01-03 03:29:10 [GPU 2] group 1/1: estimating depth map for 2/38 camera 513 (40 neighbs)...
2021-01-03 03:30:14 Agisoft Metashape Standard Version: 1.7.0 build 11736 (64 bit)
2021-01-03 03:30:14 Platform: Windows
2021-01-03 03:30:14 CPU: AMD Ryzen Threadripper 1950X 16-Core Processor (desktop)
2021-01-03 03:30:14 CPU family: 23 model: 1 signature: 800F11h
2021-01-03 03:30:14 RAM: 127.9 GB
2021-01-03 03:30:14 OpenGL Vendor: NVIDIA Corporation
2021-01-03 03:30:14 OpenGL Renderer: GeForce RTX 2080 SUPER/PCIe/SSE2
2021-01-03 03:30:14 OpenGL Version: 4.6.0 NVIDIA 460.89
2021-01-03 03:30:14 Maximum Texture Size: 32768
2021-01-03 03:30:14 Quad Buffered Stereo: not enabled
2021-01-03 03:30:14 ARB_vertex_buffer_object: supported
2021-01-03 03:30:14 ARB_texture_non_power_of_two: supported

Happy New Year!

[cbnewham]

Just to say that I've run the new version on a project with 2500 photos and it had no problems creating the depth maps or the mesh at Ultra-High.

Well done to everyone at Agisoft.

[Shadow]

Hello Shadow,

Version 1.7.0 has been published on Downloads page:
https://www.agisoft.com/downloads/installer/

Feel free to share your feedback on the latest version.

Thanks !
1.7.0 downloaded and installed, so far I haven't had any issue. Thanks for all your work !

[dcobra]

I'm noticing higher reprojection error in this release.  Have there been changes in image alignment to cause this?  On a model processed in 1.6.5 I'm getting ~0.3 pix RMS and in 1.7.0 I'm getting ~ 0.4 pix RMS.

[Alexey Pasumansky]

Hello Alexy
Same error here, metashape quit without error in the log fle, mesh set  in highest and high quality:

2021-01-03 03:24:41 Agisoft Metashape Standard Version: 1.7.0 build 11736 (64 bit)
2021-01-03 03:24:41 Platform: Windows
2021-01-03 03:24:41 CPU: AMD Ryzen Threadripper 1950X 16-Core Processor (desktop)
2021-01-03 03:24:41 CPU family: 23 model: 1 signature: 800F11h
2021-01-03 03:24:41 RAM: 127.9 GB
2021-01-03 03:24:42 OpenGL Vendor: NVIDIA Corporation
2021-01-03 03:24:42 OpenGL Renderer: GeForce RTX 2080 SUPER/PCIe/SSE2
2021-01-03 03:24:42 OpenGL Version: 4.6.0 NVIDIA 460.89
2021-01-03 03:24:42 Maximum Texture Size: 32768
2021-01-03 03:24:42 Quad Buffered Stereo: not enabled

Hello eltama,

Is the problem reproducible with only one of GPUs enabled or whether it happens with any single card enabled?

If there are any error-marked lines in Windows Event Viewer -> Windows Logs -> Application section related to metashape.exe, please export such lines to .evtx file and send it to support@agisoft.com.

[lunar]

i found this - ucrtbase.dll

- <Event xmlns="http://schemas.microsoft.com/win/2004/08/events/event">
- <System>
  <Provider Name="Application Error" />
  <EventID Qualifiers="0">1000</EventID>
  <Version>0</Version>
  <Level>2</Level>
  <Task>100</Task>
  <Opcode>0</Opcode>
  <Keywords>0x80000000000000</Keywords>
  <TimeCreated SystemTime="2021-01-02T19:05:57.7816226Z" />
  <EventRecordID>19031</EventRecordID>
  <Correlation />
  <Execution ProcessID="0" ThreadID="0" />
  <Channel>Application</Channel>
  <Computer>#########</Computer>
  <Security />
  </System>
- <EventData>
  <Data>metashape.exe</Data>
  <Data>1.0.0.1</Data>
  <Data>5fedd4ec</Data>
  <Data>ucrtbase.dll</Data>
  <Data>10.0.19041.546</Data>
  <Data>43cbc11d</Data>
  <Data>c0000409</Data>
  <Data>000000000007287e</Data>
  <Data>4744</Data>
  <Data>01d6e1320b61cad9</Data>
  <Data>C:\Program Files\Agisoft\Metashape\metashape.exe</Data>
  <Data>C:\WINDOWS\System32\ucrtbase.dll</Data>
  <Data>87f99bf8-df48-4cb3-b2ea-c3faa645f7c7</Data>
  <Data />
  <Data />
  </EventData>
  </Event>

[lunar]

 download and install 1,7 again, switch off task subdivision and have successful solution

[jrp]

Hey Alexey, can you explain what Generate Defocus Masks is?

Also very interested to know what this does and how it works.

Thanks,
JRP

[Arie]

Hey Alexey, can you explain what Generate Defocus Masks is?

Also very interested to know what this does and how it works.

Thanks,
JRP

Hi jrp,
I did some tests with this feature and it creates a mask in out-of-focus areas of the image. It seems to be similiar to Photoshops "Focus Area" tool, where based on local contrast etc. blurred parts are masked but it seems to be more sophisticated since you require a scaled model for the mask to work.
The results are great and this is super useful for creating textures without having to deal with out-of-focus areas being mapped to the texture.
This could also be useful when working with small objects where depth of field is always a problem: Create a rough, scaled model with all images, run the filter, realign with created masks (excluding feature detection on masked areas) and then running the high-res reconstruction. But I'm sure Alexey will clarify the use cases.
Thanks Agisoft team! Great update <3