Invalid value encountered in divide


Hello. I’ve been trying to process a dataset I captured with a DJI Spark, using the latest code (as of this morning). Altitude was 50 meters and the 239 images had a high degree of overlap. WebODM went through without errors or warnings, then this warning and an exit with code 1:

/code/SuperBuild/src/opensfm/opensfm/ RuntimeWarning: invalid value encountered in divide A /= s /usr/local/lib/python2.7/dist-packages/numpy/linalg/ RuntimeWarning: invalid value encountered in det r = _umath_linalg.det(a, signature=signature) Traceback (most recent call last): File “/code/SuperBuild/src/opensfm/bin/opensfm”, line 34, in <module> File “/code/SuperBuild/src/opensfm/opensfm/commands/”, line 21, in run report = reconstruction.incremental_reconstruction(data) File “/code/SuperBuild/src/opensfm/opensfm/”, line 1177, in incremental_reconstruction data, graph, reconstruction, remaining_images, gcp) File “/code/SuperBuild/src/opensfm/opensfm/”, line 1113, in grow_reconstruction align_reconstruction(reconstruction, gcp, config) File “/code/SuperBuild/src/opensfm/opensfm/”, line 20, in align_reconstruction apply_similarity(reconstruction, s, A, b) File “/code/SuperBuild/src/opensfm/opensfm/”, line 42, in apply_similarity shot.pose.set_rotation_matrix(Rp) File “/code/SuperBuild/src/opensfm/opensfm/”, line 85, in set_rotation_matrix raise ValueError(“Determinant not 1”) ValueError: Determinant not 1 Traceback (most recent call last): File “/code/”, line 47, in <module> plasm.execute(niter=1) File “/code/scripts/”, line 133, in process (context.pyopencv_path, context.opensfm_path, tree.opensfm)) File “/code/opendm/”, line 34, in run raise Exception(“Child returned {}”.format(retcode)) Exception: Child returned 1

Can you tell me what it means?


Hi :hand:, difficult to say without looking at the input data. Could you share your images (or a subset of images that cause the program to fail with the error reported above) with us by uploading the images either to Dropbox or Google Drive (or some other storage provider)?


Sure - will do.


I’ve shared my images and the task output. Would you mind having a look? To me the images look crisp and there is quite a lot of overlap. I haven’t measured it but by eye it looks like more than 60% overlap in each axis.

I’m hoping to use the Spark to teach secondary school students how to survey riparian land for an environmental assessment. I’m wondering whether there is something about the Spark camera that makes it unsuitable for map stitching. DroneDeploy had trouble with the same data set but eventually managed it. I’ve had good results on WebODM and DD using my Mavic Pro, so that’s why I wonder about the Spark.

Any thoughts about the problem would be welcome.