Splitting large data sets supported?

Dear WebODM community,
I am struggling to process my drone mission of 755 jpeg images. I believe its a memory limitation, since I only have 16GB of RAM running Windows10.

I tried one run where I set the --split and --split-overlap options in WebODM. However, it seemed not to create submodels, as suggested by this link: https://docs.opendronemap.org/large.html
It just ran on all the images, and then killed the task after a few images during the incremental reconstruction.

Defaults for most settings except for: DSM=False, DTM=True, Ignore GSD=True, Ortho and DEM pixel size=2, split =151, split-overlap=150.

Is the splitting of large data sets into submodels (to run the pipeline on every submodel after on eanother) supported? Or am I doing something wrong?

1 Like

Hello everybody,

I found the same problem. I have tried to run a set of 107 images spliting in subgroups of 20 (split = 20) and with an overlap of 50 (split-overlap = 50) but at the step of “Running local seam leveling” the process is killed due not enough RAM memory. I supposed that spliting the project into small subproject will not required to much RAM memory (because creating a task with a dataset of 20 images works fine).

Could somebody clarify me first if what I want is possible to be peformed on a Windows PC with 4 GB of RAM assigned for the docker, and second if the spliting configuration is well done?

Settings are as follow: split-overlap: 50, split: 20, dsm: true

Thanks in advance

1 Like

107 images sounds like a normal data set that will not require splitting. Splitting is for ~1e3 or ~1e4 images that need to be processed.

Also, 4GB is low on memory.

1 Like