RAM Usage Docker vs Native for ODM


Hi all,

in the past I used WebODM on Windows using Docker with about 5GB of RAM assigned to the VM.
Now I installed ODM and WebODM nativily on Ubuntu 16.04LTS. The aukerman example dataset worked.

Now I have a dataset with 92 images (DJI P3, 3000x4000 pixel) which I run with min-num-features (integer) = 5000 (default for other options).
On the ubuntu machine this consumes 16GB of RAM + all the additonal 16GB Swap Space.
The very same dataset (92 images 3000x4000, with the same parameter set) runned through the Win-Docker Version (just 5GB) with no problem.

Is that a normal behavior? During opensfm I have way more than 10 python-opensfm-processes which consumes 1-2 gig each? Any idea what might be the issue? Is it possible that this is due to some wrong configurations (?) I may did on the native ubuntu system?

Thanks for any help

btw: the example aukerman-dataset consumed about 2GB RAM overall


My guess is that your VM had less cores available for parallel processing compared to your Ubuntu machine. More cores = more memory usage for certain operations. You should try to set the opensfm-num-processes value to something lower than the default.


Thanks a lot pierotofy!

Sometimes you don’t see the forest for the trees :)…this DOES make sense.

Furthermore I discovered that there has been an additional WebODM task active which I seemed not to have terminated correctly and was not visible in WebODM frontend…so there have been 2 tasks simultaneously. Deleting the project data in the data folder did the job.

Thanks and all the best