Thanks in advance for any help with this. I am trying to figure out the options to select in order to optimize processing speed while still getting the results I want. I am using a DJI Mavic 3 Enterprise for mapping missions of Texas ranches. The only things I really care about providing my clients are orthomosaics, elevation models, and maybe plant health. Other than that, I am not interested in any other features at this time.
Currently, I am processing 1000-5000 photo sets of large ranches on an Amazon AWS EC2 server with 32 vCPU and 256 Memory (GiB). It is taking longer than I would like to do this, so I was wondering if anyone had any additional options I could select to optimize this process. Right now, my only options are: dsm: true, skip-3dmodel: true
If your scenes are largely flat, images are taken at nadir, and you have a fixed flight height above ground (stable GSD), and if you have good EXIF, you might benefit from --fast-orthophoto and --sfm-algorithm planar, though it might impact DEM product quality a bit.
I’ve been waiting for someone to run a few of these combinations against the same dataset to share the impact on time, memory, and output quality. Any chance you’ve done this?
Winner, camera calibration-accuracy:
Sift/Hahog Can’t see any significant difference
Orb Problems with camera calibration when oblique images are scarce.