Incorrect altitude on spatial view

Hey all,

As you know, with the newest platform update we can now see how Mapillary interprets our images in a 3D world. That looks awesome, but surfaced a new set of problems :slight_smile:

For instance, take a look at the following image. Unfortunately, deep linking does not support the 3D mode setting yet, so you will need to press a 3D button on the right side of your screen and select “Street + Spatial” mode: Mapillary cookie policy use

You will see that camera POVs have their GPS altitudes all over the place. Say, my latest 360 sequence you are looking at is a good 6 meters above other sequences in this photo, but actually fluctuates a lot along the sequence (i.e. previous images are “below” the base level of other’s sequences).

Firstly, I was under an impression that Mapillary ignored the altitude before. Secondly, that was one of the first 360 sequences with GoPro Fusion for me, and I am not sure if the GPS fix was that bad in terms of altitude, or that’s a usual problem for 360 sequences.

So, the question is – how can I ensure that the altitude will be correctly reflected on Mapillary? Or will that be fixed automatically by the platform? Also, should I delete the above sequence and re-visit the area afterwards?

@Mapillary : Clever trick, deriving altitude information, but going by the altitude reading of handheld Garmin GPS units like the Oregon 600, 700 and eTrex10, the altitude can be off considerably more than the horizontal error - which in town may or not place street pics somewhere inside a tall building or on the opposite side of the street. Intriguingly, the cheaper and ‘simpler’ eTrex10 provides the least inaccurate altitude.

Viewed the altitude on a recent series, near the church and market square in Haacht (Belgium): OSM indicates that first pics were taken at 11m, rising to 13 (above some datum, likely the ‘Tweede Algemene Waterpassing’ taken between 1840-1879, augmented with recent airborne surveys), the geotagged images -according to the Exiftool plug-in in XNviewMP - were taken at 20.8m, dropping to 2.7. .

Would very much prefer the option to state that the handlebar mounted action cam is at 1.2m above street level, helmet-mounted at 2.05m, and that single-frame (as opposed to time-lapse) pics are at approx. 1.8m above street level.

Would be helpful if there were an option to indicate -when taking the camera up an observation tower- that pics were taken at level 1 …5 , where level 5 = 25m above ground level.

@WASD42 : my reply would be to leave pics, provided they’re placed near where they were taken?

I’d like to see the altitude reference switched to some common source such as SRTM. Ideally this could be applied to the dataset later.