[OSM-talk] Identical Duplicate buildings

Stephan Knauss osm at stephans-server.de
Fri Jan 29 10:48:39 UTC 2021


Hello Pierre,

On 28.01.2021 23:13, Pierre Béland wrote:
> good news that you can manage to fix this in the iD editor. Discussion 
> about quality did not focus on the Editor tools.

I am sorry, if the wording caused some misunderstanding. To make it 
explicitly clear:
I did NOT do any fixes in the iD editor. I corrected OSM data by 
removing the mentioned duplicate geometries. It did affect roads or nodes.

The common part was, that IMHO it was often iD and that that the same 
user did the upload of the exact same geometry/tagging a bit later.

This was leading me and, by reading this thread, others into thinking, 
that there is a possible work-flow in which users unintentionally upload 
duplicate objects.

I use osmose to detect such issues.

If you see massively higher numbers somewhere else it might be 
interesting to see what is the common part here.
There was the suspicion that a specific usage patter triggers this, like 
early closing of the editor. But then I would expect it to happen 
everywhere with a similar probability or it concentrates on specific 
users having this pattern. I did not observe this.

Maybe it is related to bad network connectivity? This might explain why 
it is more frequent in areas which have lower network quality.

One example. This node was duplicate:
https://www.openstreetmap.org/node/6506144132/history#map=19/13.74290/100.52013

In this case the editor identified itself as "OsmAnd+ 3.3.8".


Another example:
https://www.openstreetmap.org/node/7899874464/history
Editor in this case "iD 2.18.5"

You could find them using osmose. Some areas have more of them, than 
others. In total we are at roughly 6.900. See the graph for a 
visualization of the changes over time.

http://osmose.openstreetmap.fr/en/errors/?item=1230&class=1

http://osmose.openstreetmap.fr/en/errors/graph.png?item=1230&class=1

Unless you are looking at one of the top countries, the numbers are all 
in the lower two-digit range. I recommend manually checking and fixing 
them. Often you see this way other issues which needs fixing as well.

Stephan




More information about the talk mailing list