Forum

Author Topic: Bach processing much faster than regular process  (Read 1496 times)

Alvaro L

  • Newbie
  • *
  • Posts: 40
    • View Profile
Bach processing much faster than regular process
« on: July 10, 2018, 12:09:54 PM »
Hi there

I am getting an unpredicted result since bach processing of a chunk to build a mesh is not only much faster than a regular mesh build, but also produces better results such as higher poly count, using apparently the same parameters (arbitrary, high quality) in both cases. Also in batch process there is an extra "depth maps quality" for building meshes which is not available in a regular "build mesh".

What I am doing wrong or right here?
« Last Edit: July 10, 2018, 12:11:31 PM by Alvaro L »

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14968
    • View Profile
Re: Bach processing much faster than regular process
« Reply #1 on: July 10, 2018, 03:15:35 PM »
Hello Alvaro,

Can you provide the reports files and, if possible, processing logs saved after both runs (in batch mode and in common mode)?

The new (experimental) mesh reconstruction method is described in the following article:
http://www.agisoft.com/index.php?id=48
It can be turned on via  Advanced Preferences tab.
Best regards,
Alexey Pasumansky,
Agisoft LLC

Alvaro L

  • Newbie
  • *
  • Posts: 40
    • View Profile
Re: Bach processing much faster than regular process
« Reply #2 on: July 10, 2018, 09:07:52 PM »
hi Alexey

Thanks for your answer.

I think it has to do with using depth maps to generate a mesh.  In the single process mode, if I enable visibility consistent mesh generation in preferences and reusing depth maps for mesh generation, only then I get similar results.

However, some differences still remain. The first one is that in batch process I have extra options for depth maps to generate a mesh, for instance depth maps quality. Also using depth maps in the single process mode, while I get similar render times and polycount, the mesh quality is not the same. I also get libtiff errors like the one reported here: http://www.agisoft.com/forum/index.php?topic=8689.0

Meanwhile, in batch processing the mesh quality seems higher using depth maps, and i don't get libtiff errors. So by  looking at the results, similarly named processes seem to be using completely different algorithms. But also in batch processing of mesh building there is also a "reusing depth maps" option available if I enable visibility consistent mesh generation in preferences.

It would be great to have some clarification about this. I also believe you have a little big issue with software documentation. I am posting logs only since I use photoscan standard and I have no way to generate a report.