Single point panorama

I’m taking images of charging stations for an integration in OSM. I stand in a single point and always take five pictures (-90°, -45°, 0°, 45° 90°).
Since the GPS device is not 100% accurate it is a little off all the time so I have 5 spots with wrong positions and wrong viewing angles. Is there an option to group all points at one place? The fixing the degrees is quite easy, but moving all images onto one spot is quite painful.

Nice would be if I could write a little skrip al á: A is the position for all images and the main view direction is X°. Now the script will take care of the rest.

Mapillary don’t have any features and I am not aware of a script that does exactly that. I have seen a couple of scripts that will average several GPX-files but this is quite different.

If you can write a script that would be nice. Personally I would calculate the average of lat and lon, then find the point that differs mostly from the average and then make a calculation with the remaining 4 points.

In general I would expect the last point to be most accurate, because the GPS have had the most time to lock onto additional sattelites.

quick and dirty would be to copy one gps exif to the rest of the images, if you need them in one precise position.
I am not sure why though - Mapillary’s combined panning would sort of stitch all 4 into a panorama if it finds common points, which it should. If you want to skip this step, you’d need to stitch on device (using a panorama capture mode available in the OS or an app) or in post processing

1 Like

Have you tried geosetter ?

1 Like

Or you could dive into Hugin? And try to stitch a panorama yourself?

PS: or you could buy a YI360 for only $ 160 at Amazon :stuck_out_tongue:

1 Like

And how do you test the charging stations ?

Are there any help pages on how you write a script for mapillary?

Yes, I want to have all points in one location (all images were taken from the same spot)
The compass does not seem to work very well. Mapllary will not combine or find any spots if the images are not in order (see for example and for an updated version)

See for an version. Note that one mages does a good sliding animaton while the others do not have that animation (so the algorithm did not find enough common spots even they should have a few common points) Also note that not only the GPS positions but also the viewing direction is only correct in a very general manner.

However, I want to avoid to upload the images with the webuploader. This is just another step that takes a lot of time. And since no GPS point is usually correct, I have to edit the points anyway.

@c64 I did not try that so far, but again: I would have to import the images from my phone and probably upload them via the web uploader - which takes extra time. I tried to get around that by using a gpx track on my phone, but that did not work out.

@eesger Same here with the extra step. Hugin also takes quite some time.

@filipc Why testing charging stations?

They can be idle for a long time before they are activated.

sometimes processing of common points might take a while. you could also try manually editing position and compass direction for those not matched up.
and of course, as mentioned above, using an exiftool or similar script to put them in one place with +90def offsets might help

I have a strong doubt if Mapillary recalculates the images if you correct them after uploading. In my experience, if they aren’t matched after the initial upload, they still won’t after you edit them. I think any editing should be done before uploading. By the way, uploading can theoretically be done by the same script that would edit them.

When you have everything in place so all processing is always right, then you can just upload as the final step in a script. But untill you have seen it work time and time again, it is a real pain to correct anything major after upload, so keep a manual check for some time.