Child returned 137


#1

I run ODM on a cloud cluster, 60 G RAM and 16 cores, Ubuntu 16.04.
I try to generate an orthophoto of 2700 eBee, S.O.D.A images.

only force-ccd 13.13 fails, child returned 137
–fast-orthophoto works
–orthophoto-resolution 10 works
–orthophoto–resolution 5 fails, child returned 137

are there any relevant options I can use to get around this problem?


#2

That’s a lot of images to do in one go, even with 60GB of RAM. I suggest using the (admittedly nacent) spli-merge approach: http://docs.opendronemap.org/large.html. I have processed 7K images using less than 16GB of RAM using this approach.

Can you share your log so we can see where it fails as well?


#3

Split-merge is a little messy to use at this point. You have to check out an older version of ODM: git checkout d80d0b2992c96d4a336319f46e76b844dfa33e84. This is something we’ll be sprinting on soon, with completion in February if all goes well.

What I’ve been doing is running on the above branch, then checking out the current branch, removing all odm_* directories and rerunning with the current branch. I will be writing up this process shortly.


#4

OK, I will rerun and share the log file.

Question, how to load an older ODM version; git checkout d80d0b2992c96d4a336319f46e76b844dfa33e84

Last part of error log:

Processing view view_2769.mve…
Processing view view_2770.mve…
Processing view view_2771.mve…

Created 2772 views with 2764 valid cameras.
Imported 2764 undistorted images.
[DEBUG] running /code/SuperBuild/src/elibs/smvs/app/smvsrecon -t16 -a1.0 --max-pixels=409600 -o1 --debug-lvl=0 --force /code/smvs
Shading-aware Multi-view Stereo (built on Nov 2 2018, 01:48:03)

Initializing scene with 2764 views…
Initialized 2764 views (max ID is 2771), took 1479ms.
Reading Photosynther file (2772 cameras, 1965503 features)…
Automatic input scale: 3
Input embedding: undist-L3
Output embedding: smvs-B3
Running view selection for 2764 views… done, took 5685.9s.
Resizing input images for 2764 views… Killed
Traceback (most recent call last):
File “/code/run.py”, line 47, in
plasm.execute(niter=1)
File “/code/scripts/smvs.py”, line 85, in process
system.run(’%s %s %s’ % (context.smvs_path, ’ '.join(config), tree.smvs))
File “/code/opendm/system.py”, line 34, in run
raise Exception(“Child returned {}”.format(retcode))
Exception: Child returned 137
Tue Nov 27 08:47:26 UTC 2018


#5

I have the same issue and I found no solution.
I run ODM native on Ubuntu 16.04 (no docker), with 64 G RAM and 32 cores. I tested a dataset with 13 images of 60M average each one and the process fails on orthophoto generation.

Generating texture patches:
Running… Killed
Traceback (most recent call last):
File “/home/inaoe01/ODM/run.py”, line 47, in
plasm.execute(niter=1)
File “/home/inaoe01/ODM/scripts/mvstex.py”, line 126, in process
‘-n {nadirWeight}’.format(**kwargs))
File “/home/inaoe01/ODM/opendm/system.py”, line 34, in run
raise Exception(“Child returned {}”.format(retcode))
Exception: Child returned 137

Many of the solutions given in this community are for docker users, but this is not my case.

I have been following the issue #445, that suggests using “–use-opensfm-pointcloud” or “-opensfm-processes n” flags, but my GitHub ODM version (OpenDroneMap 0.4.1) doesn’t recognize those flags. Which version of ODM do you recommend?

Thanks in advance.


#6

I am having the same issue with 572 images.
I am new to WebODM and was unsure if it was just me.

Processing Node: node-odm-1:3000 (auto)

Options: min-num-features: 12000, dem-terrain-type: FlatForest, max-concurrency: 1, rerun-from: opensfm

Last section of error code:
Running view selection for 572 views… done, took 604.308s. Resizing input images for 572 views… Killed Traceback (most recent call last): File “/code/run.py”, line 47, in <module> plasm.execute(niter=1) File “/code/scripts/smvs.py”, line 85, in process system.run(’%s %s %s’ % (context.smvs_path, ’ '.join(config), tree.smvs)) File “/code/opendm/system.py”, line 34, in run raise Exception(“Child returned {}”.format(retcode)) Exception : Child returned 137


#7

I suggest to use more memory if you have. I have run successfully 2800 images n a cloud computer with 80 G RAM and 20 Cores. You can use the orthophoto only option, --fast-orthophoto , tis skips 3d dem generation, and produces only the orthophoto.