How does setting feature-quality to different values change the results?
Does feature-quality resizes the image?
Like mentioned in the docs, whenever, we are increasing the pc-quality, processing time roughly increases by 4x. Is there any similar calculation for feature-quality as well?
Just end of last year I did some benchmarking of different settings and hardware.
Yes it does, except on ultra. There is some documentation around where it states the sizes used, when choosing different quality settings. As far as I remember the scaling is something like ultra=1.0, high=0.75 and middle like 0.5 of input picture size.
But that does not effect your GSD or Orthophoto resolution, afaik.
To my experience feature-extraction on ultra does not increase the computing time a lot. It mostly increased the number of detected features and therefore increased the robustness of the reconstruction.
For the projects I run (500-3000 images, 12MP) I mainly use ultra as feature-extraction quality.
In the topic below you can find some comparisons between high and ultra settings and how a GPU effects the results:
Correct, it resizes the images only for/during feature detection/extraction, but this does not affect the images used for creating the Orthophoto and texturing the mesh.
Just read it. Loved your analysis.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.