[Imports] Bing Building Import

Greg Morgan dr.kludge.gm at gmail.com
Wed Jul 4 17:30:40 UTC 2018


On Wed, Jul 4, 2018 at 1:34 AM, Christoph Hormann <osm at imagico.de> wrote:

>
> Greg,
>
> I will comment on a few of the new things you have written but like to
> emphasize this is still not an import review because a lot of
> information required for that is missing.  You could however read up
> old discussions on previous building imports here to get an idea on the

requirements and suggestions made for those.
>

Got it.  It is not a review but a helpful number of ideas.  Thank you for
your time.


> > [...]  In my early opinion, the foot prints are no
> > better nor no worse than a craft mapper,s drawing
>



> This is always a pretty meaningless comparison because it is apples and
> peaches.  When i talk about "quality aspects" i mean quantifiable
> measures of quality.
>
> > [...]  As your
> > other post provided an idea of starting with Montana, that will not
> > be useful in my case.
>
> I suggested rural Montana might be a good place to start if you
> intend "to poke the data for quality issues".  Since that is not what
> you want to do my suggestion is pretty useless for you obviously.
>
> > > Microsoft's process documentation contains a number of hints that
> > > indicate things can go wrong in the process in ways that are likely
> > > to produce significant errors of kinds that are very unlikely to
> > > happen in manual mapping.  Without having reliable data on how
> > > often these things do happen (and how this varies between different
> > > geographic settings) you would essentially be doing a blind import.
> >
>

Christoph, as an analogy you have read that you can get sun burnt if you go
outside.  Now you are afraid to go outside because you read about the
resulting skin cancer. Microsoft is just saying that the data us good but
use at your own risk.  They also point out that we should go through the
import process.  Keeping up with the analogy, I have been sun burnt in
Montana, Wyoming, Utah, Colorado, California, and Arizona for sure.  There
are other states where I have been sun burnt but they are too numerous to
list.  My point is that I understand what is rural America in all the
states listed.  The sample I have provided so far is in rural Arizona.  It
is not in metro Phoenix.  I can parse Montana but that would not add any
any more detail to this discussion.  The test.osm file here is rural
America but in Arizona by the border with Mexico.  People who live in this
area are either retires or work for the border patrol. https://drive.google.
com/open?id=1I7BPMKLgABk8ikUdEPFpl6zKgh9E-sDN



> > Depending on the craft mapper, hand drawn buildings can have the same
> > problems. [...]
>
> As i have been very clear about this is not the case:
>
> > > Microsoft's process documentation contains a number of hints that
> > > indicate things can go wrong in the process in ways that are likely
> > > to produce significant errors of kinds that are very unlikely to
> > > happen in manual mapping.
>
> The only thing that could convince me to change this assessment would
> be - as already mentioned - a thorough analysis of the quality that
> holds up to scientific scrutiny.  And it is frankly much more likely
> that such analysis would confirm my impression.  If it does not that
> would mean Microsoft has made progress in the field that absolutely
> dwarfs everyone else working in the area and if that was the case we
> would see it on the stock market. ;-)
>
> Don't make the mistake of assuming this to be a building data set like
> various ones produced by local authorities or mapping institutions some
> of which have been considered for import or have been imported in the
> past.  It is not.  Therefore again my statement: IMO this means that a
> proper import review would only be possible based on a thorough
> analysis of the quality of Microsoft's product that holds up to
> scientific scrutiny.
>

Continuing on with the sun burnt  analogy, if I wear a high blocking sun
screen I will not get burnt as much.  I could careless about the mapper's
assessment from Alaska.  His imagery is leaf on compared to my imagery
which is leaf off.  Because I live in a desert, I will have fewer issues
than the Alaskan mapper's data set.  There are no trees hanging over the
buildings in question.  Here's what I have done so far to protect myself
from claims on both sides i.e. the data is good and the data is bad.

* So I see that there will need to be some checking done as part of the
import process. Here are some early observations of the CNTK process that
MS used with 32 GPUs.

* The we have many solar panels in Arizona because of a state mandate. Many
of these are used as covered parking.  The tag of yes would have to change
to roof.  That can happen during the import or as a second QA pass.

* The Q in JOSM will be required for some of the black topped roofs or
solar panels.

* It is curios that some of the black roof or darker colored roofs are
missed.  When I craft map, sometimes I get interrupted, sometimes I get
bored and I never get back to the missing buildings.   Oh well!  Either
another mapper comes along and picks up the missing work or sits for
awhile.  That is what I mean by the data is no better nor no worse than a
normal volunteer's work flow.

*  All the round buildings are square.  That may be the result of CNTK
making sure that buildings are square. The work around is to add a few more
nodes and hit the O key in Josm

* I noticed one or two buildings where they included a concrete slab as
part of the building.  The X tool in JOSM will be used to scale back that
part of the building.  Again, I've had to clean up these kinds of issues
with other craft mappers including myself.  Buildings that I mapped from
the prior version of Bing or Yahoo, are clearer with the last Bing imagery
release in my area.

* The imagery in this area of the test.osm file is satellite based. It is
not of the same high quality .5 meter and below I get to use in the metro
Phoenix area.  Moreover, the metro Phoenix are is flown imagery so most of
it is very high nadir.  MS states that they have a number of sources of
imagery providers.  The buildings still look respectable.

* There was one school that I was puzzled why it came out the way it did.
This is a non issue because there is an existing building.  The Bing
building will be deleted.

I have to dash.  Perhaps I can post some screen shots of the test.osm
file.  What I have done so far is to have two layers in Josm to compare the
data with the existing OSM data.  The test file is here
https://drive.google.com/file/d/1I7BPMKLgABk8ikUdEPFpl6zKgh9E-sDN/view .
The whole Google drive folder is here
https://drive.google.com/open?id=1FhQd8fSIbx9OyG6vFAaKkBfYn10Rw0Da .  The
information may be useful to other mappers considering the Bing
footprints.  Arizona is one of the larger data files.  I can parse other
states if that will be helpful to other US Mappers.  And yes I was going to
looke at both Wyoming and Montana.

Regards,
Greg
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20180704/d953097d/attachment-0001.html>


More information about the Imports mailing list