[Talk-GB] building shapes from OS Street View

Dave Stubbs osm.list at randomjunk.co.uk
Tue Apr 6 18:07:41 BST 2010


On Tue, Apr 6, 2010 at 5:26 PM, TimSC <mapping at sheerman-chase.org.uk> wrote:
> Andy Allan wrote:
>> On Tue, Apr 6, 2010 at 12:46 PM, TimSC <mapping at sheerman-chase.org.uk> wrote:
>>
>>> If we use only manual surveying, we can only
>>> achieve coverage of about 1%. I don't think that is satisfactory. Imports
>>> are therefore very much appropriate for buildings.
>>>
>>
>> You're missing the point on several levels. For buildings it's quite
>> possible to trace the outlines from aerial imagery, where we have it.
>> And in some parts of the country, we have better imagery than the
>> building outlines shown on Street View - yahoo, and soon the Surrey
>> imagery, for starters. So it's not a case of Street View or nothing. I
>> hope you have seen the vast tracts of London that have building
>> outlines in OSM already - all of much higher detail than Street View?
>>
> It's not vast, it doesn't even get to the edge of the underground zone
> 1. And the detail is comparable to my eye, except for many omissions in
> OSM. I am not sure how you are quantifying quality. I suspect that the
> OSM data is more up to date (even if Yahoo is a few years out of date),
> but OSM still very incomplete even in this supposedly well mapped area.
> I suspect with the few OSM contributors doing tracing the various
> sources and given their limited coverage, we are still looking like
> converging on poor overall coverage given years of effort, so my
> original point still stands. I say we can compliment our other sources
> with automatic tracing (be that by importing or editor tools).
>
> I guess the question is how much progress can we make on building
> outlines over different time scales, given different approaches?
>> If you are not talking
>> about "bulk imports" then please don't call your ideas imports,
>> otherwise you confuse people as to your intentions.
> I am discussing automatic tracing which applies to both editor tools and
> imports. There is no rule that I have to discuss one option exclusively.
> But I was leaning towards more the import paradigm, while the majority
> seems to be for editor tools. Andy, from my perspective, you have not
> given a single justified reason against doing imports, so I can't really
> rebut your position (although other people have made valid points). I
> suggest you get a bit more constructive and outline your vision for the
> way ahead on this issue? Continue, as is, with Yahoo and so on?
>

The thing we're looking to not have is automatic imports. London has a
lot of buildings, but only in certain areas. What I don't want to have
to do is wake up one morning to discover someone has helpfully
imported auto-detected rectangles over the top, meaning I have to
spend the next three days/years cleaning up the data. If all you want
to do is load some buildings into an area (however that's
implemented), fix it up to avoid duplicates and conflicts with
existing data such as roads, and upload that, then fine. If you want
to spend the time comparing against other sources too then even
better.

If you want to dump buildings into the entire country and hope that
everybody gets around to fixing all the problems in their area
afterwards, then please don't.

Basically: please don't break the map :-)
Other than that, tracing OS street view is by far the best source of
building outlines we will have in much of the UK at the moment.

Dave




More information about the Talk-GB mailing list