Support Website Contact Support Blog

Orientation of models with Iphone 12 pro

Hi folks,
do you know if models acquired with the Iphone 12 Pro and processed via Pix4Dcloud are supposed to be properly oriented? We have carried out a test on a 100 m wide scene using 300 photos and, apart from the huge time required for processing (5 days, but we suppose it is due to our trial licence), the model is not properly oriented: the true north is actually pointing to NE.

Thanks in advance

Stefano and Eugenio

Hi Stefano and Eugenio,
The long processing times were due to an internal issue, not to the type of the license you have. I have reached out to you via ticket system regarding that issue. Your single projects took at maximum 1h 6 minutes to process after the issue was fixed.
Could you tell us more about the orientation issue you are having? Which project has this issue?

Hi Stefano and Eugenio
Thanks for reaching out. Pix4Dcatch uses the phone’s internal compass and there are chances that the compass wasn’t working well. This can happen if your phone was in contact or close to a magnetic field.
Could you perhaps recalibrate the compass and try again? See how to here

Thanks and kind regards

Dear Alice, thanks for your response.

In all the three models the horizontal plane is almost perfectly horizontal, whereas the north is just random, with error spanning from 30 to 100°

To respond also to Amritha Narayanan, for each site before photo acquisition we have checked the Iphone internal compass against an analogic compass and the error is in the order of a couple of degree.

Stefano and Eugenio

Hi Stefano and Eugenio,

Thanks for your feedback again. The depth information and the orientations values are made available via ARKit, which seems not to follow exactly the device compass. We will internally evaluate the problem and elaborate on some possible solutions to improve the absolute orientation of the models.


Ciao Alice,
since it seems that the misorientation only affects the rotation about the Z axis, a viable solution could be that of running a final check, by comparing the reconstructed and measured directions of each image and, in case of a remarkable discrepancy, to perform a final rotation about the Z axis.

Stefano and Eugenio