I have been working on the performance of ODM in multi-core super computing environments, and we have been interested in the fact whether split/merge could be parallelized. This would enable to use more cores, thus enhancing the processing of images (at the cost of memory of course)
Yes. See ClusterODM and if you’re in a slurm environment, recent pull requests by Nghi: Pull requests · OpenDroneMap/ClusterODM · GitHub
Regardless, ClusterODM with split will handle node allocations for you and allow you parallel node use at the most computationally intensive stages.