[Talk-ca] Importing buildings in Canada

Tim Elrick osm at elrick.de
Fri Jan 17 17:12:06 UTC 2020


Hi John,

As stated before, I don't consider the Microsoft dataset being close to 
the minimum quality requirements I would expect from any automated 
building entry into OSM. If you just want to display buildings, you can 
download the MS dataset and use it right away - no need to import into 
OSM. I think, the MS dataset has value as proof of concept and to count 
the number of buildings in a given area (e.g. to estimate market size 
for roofers, estimate number of persons living there for desaster 
relief, etc.). I also think, when Microsoft feeds its algorithm with 
higher resolution data than they did (I don't recall, but I think they 
only used the regular Bing data) they will probably end up with building 
footprints that will meet our/my quality requirements for import into 
OSM one day.

For me, the value of OSM is having accurate information in terms of tags 
and geometry. Otherwise, we could join Wikimapia; they don't care too 
much about geometry accuracy but emphasize on content/tags of objects. 
Pretty interesting project, but different from OSM.

Cheers,
Tim

On 2020-01-17 10:40, john whelan wrote:
  >first, to add missing buildings (if it were
just for this purpose we could also use the much bigger Microsoft
dataset)

I can't resist.  Does this infer that for parts of the country without
Stat Can data we are happy to import Microsoft dataset buildings as is?
Or would we wish to wait until we have some more imports done before
looking at preprocessing them in some way first.

Thanks John



On Thu, Jan 16, 2020, 10:11 PM Tim Elrick, <osm at elrick.de
<mailto:osm at elrick.de>> wrote:

     I would assume in most cases the imported building footprint will be
     more precise than existing data. For me, this would be a reason to
     replace already existing objects. However, I think this is a case by
     case decision. However, I think it is important to keep tags and
     history
     of buildings already existent in OSM. This is how I would
     read/interpret
     the import guideline stated by Nate: "If you are importing data where
     there is already some data in OSM, then *you need to combine this data*
     in an appropriate way or suppress the import of features with overlap
     with existing data." (emphasis added by me)

     However, that just means, the import, hence, is nothing easy and could
     not be achieve quickly, I would assume. One way of making sure that
     this
     is dealt with diligently, would be setting the tasking manager to
     'experienced mappers only'. We would have to ask James, who is in
     charge
     of the Canada Tasking Manager, how to edit/set up the 'experienced
     mapper role' in the TM. It might be possible to feed in a list of
     mappers manually or to set a threshold of objects/changesets that they
     must have entered in OSM. However, maybe only mappers who feel
     experienced enough to handle the import would contribute to the TM
     project anyway and we let everyone judge on their own and don't
     restrict
     access.

     If we were to separate the new and overlapping buildings, I am also
     leaning towards Daniel's assessment. I would be afraid to cause more
     issues than by doing it all at once (with a reasonable tile size, of
     course).

     In the end, the main point of importing this specific dataset fulfils
     two purposes, in my opinion: first, to add missing buildings (if it
     were
     just for this purpose we could also use the much bigger Microsoft
     dataset), second, to get the best geospatial representation possible in
     our OSM database. That means, we defer from using the Microsoft dataset
     and use the much higher quality data from the ODB. This also means that
     we should replace already existing buildings (yet keeping tags and
     history) wherever the ODB footprint is more precise than the
     existing one.

     Just my two cents here,
     Tim

     _______________________________________________
     Talk-ca mailing list
     Talk-ca at openstreetmap.org <mailto:Talk-ca at openstreetmap.org>
     https://lists.openstreetmap.org/listinfo/talk-ca




More information about the Talk-ca mailing list