Image heading data broken

I think I’ve been having this issue for over half a year now, and it makes phone capture impossible or frustrating.

Previously, before the gpx update and even a couple of versions prior to that, the only issues with image headings I would have would be due to the magnetometer itself being wonky, metal interference or old calibration. This was quite rare and I could just walk around in Manual mode turning the phone in different directions and the photos would almost always have the correct heading.

Then something happened, and image heading got absolutely broken. It would show correctly during capture (presumably due to iOS compass data being shows), but then the actual heading on upload would be 90-100deg off. I’ve tried debugging this together with mapillary support, and you might remember iOS app releases claiming to fix the issues.

Decided to try out the latest version today, and indeed there doesn’t seem to be that much of an error, however the heading is still +/- 40deg on some images. On some, it would seem it is interpolated from the gpx track rather then the compass. An image captured in the same position could have heading variances of 30-40deg, with the phone held still.

Am I the only one having this?

yo tengo el mismo problema pues cuando tienes mas de dos camaras con diferente rumbo y asignas la misma cordenada algunas camaras pierden el rumbo correcto

“I have the same problem because when you have more than two cameras with different directions and you assign the same string some cameras lose the right direction” courtesy of an online translator

1 Like

I might have the same problem - see Mapillary for an example.

https://github.com/mapillary/mapillary_issues/issues/1451 suggested problems with heading/compass in a car, but the sequence above was not in a car.

I think that old issue was relevant in 2015 when compass values were (apparently) computed from reconstruction. It would seem they are now interpolated (poorly)

Should a new [bugreport] thread be used then?
My phone sequences are done with iPhone 6, iOS 12.latest.

I have had this bug on 6s and X, both running 11, 12 and later 13. Maybe magnetometers are defective in mine.
Not sure if anyone from the dev team looks through the threads, [bugreport] or not