Task Settings Seem to Make Image GPS Locations Go Crazy?

[ludeykrus]

Good Day,

I have been experimenting with settings now that I have 256gb of RAM under the hood of my workstation, and have ran into some interesting results after tweaking the settings in a way that I expected to produce a better result. This is regarding one such instance.

I am running the newest WebODM version through Docker under Hyper-V on Windows 10 Pro. Workstation has ~220gb of RAM and 1.3TB of hard drive space allocated to Docker/Webodm. Can upload the images, but would prefer not to publicly post them at this time. Dropbox link is below; the files included are as follows:

Dropbox - File Deleted

-Settings - shows settings for a number of tasks ran for the same dataset with differing settings. All projects produced a good ortho/result with good GPS locations except for the bottom. “Screwy” indicates the ‘bad’ project task, and ‘Good’ indicates a proper result from the top task on this list.
-Good_Map and Good_Cameras - the resulting ortho and camera GPS locations from the top task
-Screwy_Map and Screwy_Cameras - the resulting ortho and camera GPS locations from the bottom ‘bad’ task.
-Logs_Screwy and Logs_Good - the console logs from each respective project.

I’m hoping someone may be able to help me figure out which setting is causing the issues I’m seeing. It appears that something with the bottom task’s settings is causing the GPS locations of each image on the right side of the screen to go crazy, which appears to result in no data in those areas. Let me know if additional information is needed.

[Saijin_Naib]

Thanks for the great information and the logs/etc.

Hopefully we can get some clarity on what is actually happening here, but from what I can see from the Screwy project/logs, nothing jumps out as amiss so far.

Could you screenshot us the version information from your processing nodes?

[ludeykrus]

Sure thing:

[Saijin_Naib]

Can you update your OpenDroneMap from docker? I think there should be a v2.4.4 right now. Not sure if it’ll fix this behavior, but worth trying.

[ludeykrus]

I thought I was 100% up to date on this computer, but turns out I had updated my other two and this was still a version behind! Updated through Git Bash. Reran the troublesome settings, and all appears to be right in the world?! Which is super weird, I changed some settings slightly on the original dataset and got similar or worse results (min features, octree depth, mesh size, and ortho cutline if i remember right). Have not reset the computer during. Perhaps it was a fluke?

I was also running some snow-day datasets through and noticed when I ran the above-listed ‘improvements’ in settings, I got much worse results. The best results occurred when I used the ‘high resolution’ setting and tweaked the resolutions down to 0.5 inches, my shot GSD. I’ll have to rerun those with the tweaked settings to see if their results improve as well…