Author Topic: 8192x8192 generated texture has the same level of detail as 4096x4096 texture?  (Read 3941 times)


  • Newbie
  • *
  • Posts: 2
    • View Profile
Hi guys,

We've been using Agisoft a lot recently to scan entire rooms. We were generated the rooms' textures at 4096x4096 (single texture), and we figured that we could potentially increase the detail of detail by generating 8192x8192 textures instead.

So we tried doing this, but didn't notice any visual difference or increase in texture detail in the final model.
To be certain, I took the previously generated 4096x4096 texture and upscaled it to 8192x8192 in photoshop, and it looked *exactly* the same as the 8192x8192 texture generated by Agisoft.
When generating 8192x8192 textures, does Agisoft first generate a 4096x4096 and then just upscale to 8192x8192? Or is there some other reason this is occurring?
Our photographs should be high enough detail to generate good 8192x8192 textures, so I am not sure why this is happening. Are there any other settings that need to be adjusted?

Has anyone else noticed the same problem? Or rather, is 8192x8192 actually giving increased detail for anyone vs 4096x4096?
We're building textures using mapping mode Generic and blending mode Mosaic.

Version: Agisoft PhotoScan Standard 1.3.4 build 5067 (64 bit)

We would like to increase our texture quality so any help would be appreciated.

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14854
    • View Profile
Hello matpro,

Can you share both texture files generated in 8192 and 4096 resolution for the same project, using Keep UV option?
Best regards,
Alexey Pasumansky,
Agisoft LLC


  • Newbie
  • *
  • Posts: 2
    • View Profile
Hi Alexey,

I tried putting together some sample data, although the surface area is significantly smaller (I couldn't share our previous project data that we were having trouble with).


This time, it seems like there might be a slight difference between the two resolutions, although it's pretty hard to tell. What do you think?

I feel like there should be more of an improvement between 4086 and 8192, but it's also possible that the amount of surface area for this test was too small, so instead I reached a limit with the detail/quality level of each photograph - I'm not really sure. If this data isn't sufficient, I think I should photograph a much larger surface area so that the pixel/meter density becomes the definite bottleneck. I might be able to do some more tests and try to make better sample data, but I'm not sure when I'll be able to get to that.