Running ODM on a research computational cluster

Hello,

I am interested in installing and using OpenDroneMap on a large research cluster. I can access multiple nodes and many hundreds of cores if need be. I can also access nodes equipped with GPUs.

Our interest is in obtaining detailed orthomosaics and elevation models of nearly dry river beds. The area of interest will be fairly large (possible > 1 km2), hence our desire to use a computation cluster.

If somebody could please provide information on the best way to achieve this it would be greatly appreciated.

Thank you

Jason

1 Like

Welcome, Jason!

Really interesting question and research.

I think, if your cluster is running Linux/Docker, the easiest route would likely be the GPU-enabled Docker images for odm and at least one marshaller with cluster-odm.

See the GPU-enabled ODM docker image here:
https://hub.docker.com/layers/opendronemap/odm/gpu/images/sha256-f3a1f689b914f2028dcbb1c4352ed5f14b1757c7ee9e102d5152259e778cc82b?context=explore

Here’s the documentation on the split/merge pipeline:
https://docs.opendronemap.org/large/

I have run on 10s of square kms, so it’s feasible for sure. Do you have a target elevation model resolution?

If you’re running on a cluster, is it an HPC architecture?

Also: welcome!

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.