Error processing - ClusterODM

Hey guys

I’m trying to process about 646 photos locally using ClusterODM with two computers I have. I think I’ve set everything up correctly, but it keeps failing. In the dropbox link attached is all of the photos, the console output which will show the error, both computer’s hardware info and the docker resources info. Any help would be appreciated!!

Hi @John_Provost - can you successfully run a smaller dataset (e.g., Garfield or Toledo from here) ?

@coreysnipes yes I’ve ran multiple data sets previous to this. None over 500 images though. This is also my first time using ClusterODM and split-merge

OK, good deal. How about running a smaller set through cluster/split-merge? That would help rule out some variables such as - is it the cluster, or the dataset?

@coreysnipes I just ran the dataset “Garfield” from the link you suggested above with the same settings I used for ClusterODM and split-merge. The only difference being the amount of photos. I still had it split it up into 3 separate sub-models. It completed in 13min. So I assume that means it has something to do with the dataset?

Well after 9 hours it eventually did process by changing some options around. The change that seemed to make the biggest difference was the split size. I upped it to 300 images per split. Could this have fixed it because I don’t have enough physical ram to process too many submodels?

Also I remember reading that if you use clusterODM it will not make a 3D model. From this test, it seems to be true, but I wanted to confirm this.

Smaller splits can sometimes be tricky because they might not have good structure. This could be why. :thinking:

As far as 3D models, you are correct.

2 Likes

John, sorry for the delay in responding, and thank you for reporting your final outcome. I am just getting started with cluster and split/merge myself and don’t know much in that area. Glad to see @smathermather jumped in with some info.

1 Like