[Talk-GB] Google maps added addresses!

SK53 sk53.osm at gmail.com
Tue Dec 29 17:16:06 UTC 2020

Whilst I share this
<https://demo.f4map.com/#lat=52.9744324&lon=-1.1972836&zoom=18> particular
frustration I don't think any of the 3 popular approaches (standalone,
entrance, building) is perfect, each has downsides for specific uses.
Standalone suffers because it may not be easy to correctly attach an
address to other elements which use that address. The OSM data model also
militates treating addresses as first-class objects (but in the real-world
most systems & applications don't either).

As discussion earlier on terraced houses has raised, there may be
advantages to retaining the shell of the entire terrace as a separate
building despite validators strong objections to buildings within
buildings. There are clearly properties which belong to the terrace (roof
shape, name, listed building status
<https://www.openstreetmap.org/way/836645301> & so forth) which cannot be
readily split across all the houses in the terrace and properties which are
distinct to each house (roof:material, fake stone cladding, uprn & so on).

For some years I've been muttering about how to discover 'emergent'
properties of things automatically from what actually gets mapped on OSM
(examples urban area
river & glacier flowlines
<http://sk53-osm.blogspot.com/2016/10/skeletons-in-water.html>, river
basins and their watersheds, avenues of trees where trees are mapped
individually, etc). Terraces & semi-detached houses would appear to be
relatively trivial examples, but the discussion so far suggests that it may
be hard to computationally derive all features of the terrace.


On Tue, 29 Dec 2020 at 10:27, Michael Collinson <mike at ayeltd.biz> wrote:

> +++1 to mapping addresses as standalone nodes!
> While it is an alternative accepted style to put them on buildings, I
> personally loathe it. A building is a building, an address is an address. A
> one-to-one relation is common but not 100%. The address may not even be
> logically applicable to a building, examples: schools (a cluster of
> buildings on a site), recreation grounds (perhaps no building at all).
> Dividing a building vertically for the purpose of address in messy. Yes,
> there is some logic to the concept of separate living/working volumes but
> then that should apply to horizontal division as will, common in some parts
> of the world. For me, a semi-detached house is a semi-detached house and a
> terrace is a terrace.
> And the worst thing, and my main motivation for writing, is that is
> destroys simple 3-D modelling,
> https://wiki.openstreetmap.org/wiki/Simple_3D_buildings . A semi-detached
> house with a hipped roof becomes impossible to model if two conjoined
> buildings or even building parts as far as I can see (?).
> I suggest always modelling addresses as separate nodes placed in a logical
> place for navigation.
> A jolly Christmas rant brought to you by Mike. Best wishes to all.
> On 2020-12-29 09:28, Simon Poole wrote:
> At the danger of pointing out the blatantly obvious: you can easily survey
> and add addresses as stand alone nodes without adding buildings before.
> It is quite a fast process and, at least for me, is only limited by
> walking speed and getting distracted by other details which you tend to
> only see when surveying on foot. It is what we used to do before aerial
> imagery was widely available and will result in fully functional routing.
> Given how painful correcting building geometries is I would always prefer
> an address node over mapping a building outline from a sub-par source.
> Simon
> Am 28.12.2020 um 17:21 schrieb Rob Nickerson:
> Hi all,
> I just spotted that Google Maps has added house numbers to their map of
> the UK. They are all over the place - does anyone recognise the source?
> What can we do to improve coverage of addresses in OSM? I notice that we
> have some pretty good aerial imagery now. Should we see if we can get good
> building outline from an AI / machine learning approach? If the quality is
> good we can then use these to help add addresses. For example we can ask
> new mappers to add addresses using tools such as StreetComplete.
> Any thoughts much appreciated. I have a feeling that if we can come up
> with a plan we may be able to get some help from several of the big tech
> companies now interested in the UK. It would be better if we were steering
> this rather than it happening to us.
> Best wishes,
> *Rob*
> _______________________________________________
> Talk-GB mailing listTalk-GB at openstreetmap.orghttps://lists.openstreetmap.org/listinfo/talk-gb
> _______________________________________________
> Talk-GB mailing listTalk-GB at openstreetmap.orghttps://lists.openstreetmap.org/listinfo/talk-gb
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20201229/73442ed6/attachment.htm>

More information about the Talk-GB mailing list