[Talk-ca] Microsoft has released its building outlines for Canada

[john whelan]

Please read Tim's comments below I suspect he meant to cc talk-ca.

My thoughts run along the same lines as Tim's basically local groups have
to look at what is available then decide for themselves how they wish to
proceed.

We have some experience in the import process available but I think it
needs to be driven by local mappers since they are the authority.

Cheerio John

On Sat, Mar 2, 2019, 7:45 PM Tim Elrick, <tim at elrick.de> wrote:

> Hi everyone,
>
> Thanks John for pointing out the new dataset!
>
> Steve, John just indicated that there is this new dataset available now. I
> am confident, that after our discussion on importing building footprints in
> Canada, the OSMappers who want to go forward with it, will provide the
> information on the import plan/wiki.
>
> You can download the Microsoft file at the site indicated by John. The
> license information can be found there as well (as mentioned by James it is
> ODbL. It is much more extensive than the Open Building Database release by
> StatCan (> 12 million buildings vs. 4.4 million buildings).
>
> So, I got interested and I had a look at the data for Montreal:
> https://imgur.com/a/PwuZ3Q5
> I chose an area where OSM data existed. As you can see OSM hand-drawn
> buildings from Bing imagery are nicely drawn and separated.
> When you compare the OSM data to the Open Building Database (OBD) data
> from StatCan (brown in my images) which draws on the données ouvertes de
> Ville de Montréal (extracted from photogrammetric imagery) you can see that
> the OBD data is somehow more precise than the OSM/Bing data, but the huge
> draw back is that the OBD data only captures the building outlines of
> attached/terraced buildings, i.e. of building blocks without separating
> single buildings (as already mentioned by me in a separated e-mail to the
> list - this is why we will have to do some pre-processing of the data).
> When you then compare this data set to the Microsoft data set (pink in my
> images), you see that their deep neural network approach fails in areas
> with terraced houses big time. Same goes for the the city centre of
> Montreal (not shown in my images).
>
> So, the Microsoft data set might work for areas with single homes and that
> would be helpful to fill in the blanks in remote areas, but for areas where
> we have data from the ODB, importing from the ODB might be better. However,
> caution has to be taken with ODB as well, as every municipality that
> contributed data might have contributed a different data source, where the
> quality should be check each time.
>
> Just my two cents here.
>
> Tim
>
>
> On 2019-03-02 17:40, john whelan wrote:
> Why are you planning to import it?
>
> Cheerio John
>
> On Sat, Mar 2, 2019, 5:26 PM OSM Volunteer stevea, <
> steveaOSM at softworkers.com> wrote:
>
>> A responsible complement to this would be a link to license information,
>> a wiki page about these data, and perhaps an Import Plan should those data
>> actually be asserted to be worthy of being responsibly imported into OSM.
>>
>> SteveA
>> California
>>
>> > On Mar 2, 2019, at 2:17 PM, john whelan <jwhelan0112 at gmail.com> wrote:
>> >
>> > https://github.com/Microsoft/CanadianBuildingFootprints
>> >
>> > So now there are two Open Data sources for building outlines in Canada.
>> >
>> > Cheerio John
>> > _______________________________________________
>> > Talk-ca mailing list
>> > Talk-ca at openstreetmap.org
>> > https://lists.openstreetmap.org/listinfo/talk-ca
>>
>>
> _______________________________________________
> Talk-ca mailing listTalk-ca at openstreetmap.orghttps://lists.openstreetmap.org/listinfo/talk-ca
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-ca/attachments/20190302/f29baa32/attachment-0001.html>