Exit code 127


#1

Hi,

new install of WebODM on win10. It seemed to go fine, I even got a small fast mosaic to work easily. But after that nothing processes and I get: Process exited with code 127. I tried the basic reboots of both WebODM and the whole computer. Tried other images. I cannot even reprocess the data that worked the first time. Attached is 1 image, new users can only upload 1. It’s Phantom 3 Pro imagery so pretty standard, Why has it worked once then just stopped?
Thanks for the help
Patrice

DJI_0267.


#2

Hey @PCdurham, could you paste the task output? It’s the text within this box:

image


#3

Possibly not related to this problem, but I highly suggest doing camera calibration with Phantom images if you haven’t already: https://github.com/dakotabenjamin/CameraCalibration


#4

I’ve worked a lot on camera calibration and whilst the theory does suggest it’s a good thing, the implementation is not so simple. The main issue is that calibration algorithms can converge on equifinal but different solutions. That means that the algorithms can find many answers. It’s sometimes called parameter correlation. If you’re interested, here’s a link to the open version (free to download and use) of a paper I wrote on the topic, with Phantom 3 data: http://dro.dur.ac.uk/20102/. Bottom line is that flying with oblique view imagery can allow for a proper calibration from the self-calibration algorithms. It’s the nadir only flight pattern that is really difficult to calibrate and results in the doming effect notorious to all SfM.

Guess I should say, I’m an academic that works a lot on drones (been flynig 11 years!) and SfM-photogrammetry, especially for river science applications. I’m trying to convert all my software to open source, so hoping ODM is getting ready to replace Photoscan…

Patrice


#5

The text box is blank :frowning: I just get the 127 exit code message on the top right like this:
image

P


#6

@PCdurham – while you are correct in principle, we do not enable by default enough parameters to handle barrel distortion in OpenSfM. This is intentional – it reduces the ambiguities in self calibration by reducing the dimensionality of solutions, but the trade-off is that there is no correction for barrel distortion. Reconstruction of phantom 3 data is going to be really poor without that correction, hence the recommendation to pre-calibrate, even with all the instability associated with that approach. I hope this helps.


#7
  1. What browser are you using
  2. Can you try another one to help isolate the origin of the issue.

#8

That’s interesting. So which parameters are left in the distortion model and what is calibrated? Other problems solved and related to the install.

Thanks
Patrice