Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Graeme Fitzpatrick
On Wed, 21 Oct 2020 at 14:53, Daniel O'Connor 
wrote:

> There's a link on the github repo to download a GeoJSON file. They haven't
> made it available via rapid yet.
>

That would probably explain why I couldn't find it!

Looking at it via JOSM,
>

& seeing that I don't JOSM, I won't worry to much just now

Thanks

Graeme
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Graeme Fitzpatrick
Where are you all finding this info?

I've had a look at RapiD & they either don't have much on the GC, or I'm
just doing something wrong (which is quite, quite possible! :-)), because I
can't see anything to check?

Thanks

Graeme
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Daniel O'Connor
Looks like my other message didn't go through.

Agree re the slight rotational issues; and it being at about the level of a
new-intermediate contributor from the samples around Adelaide I managed to
look at.

I found sometimes, it would mistake bright concrete as part of a building
footprint.

Because of the simple way I split the files up, I've only got sporadic
coverage - a few buildings per street from my 10% sample. I'm not sure if
the original file is sorted by longitude/latitude in any way; or if the
model simply missed many of these. @Andrew Harvey  -
can you shed some light on this?

On Wed, Oct 21, 2020 at 12:30 PM Andrew Harvey 
wrote:

> Also buildings which are touching like a shop fronts are just one polygon,
> but I don't think you'd ever be able to do this too reliably without a
> survey anyway. Even if there is a small gap between buildings eg a garage
> and residence sometimes it will join them into a single polygon.
>
> What I feel it does do well is actually identify buildings, I haven't
> found any false positives, only a few false negatives where a building is
> hidden under dense tree coverage.
>
> On Wed, 21 Oct 2020 at 12:52, Andrew Harvey 
> wrote:
>
>> On Tue, 20 Oct 2020 at 21:20, Simon Poole  wrote:
>>
>>> Just as a comment: there is nothing so time consuming as fixing badly
>>> mapped buildings (essentially drawing them from scratch is nearly always
>>> faster), I would only import building outlines that are at a quality level
>>> that you would not want to change them except if the building itself has
>>> been modified.
>>>
>>
>> I took a look at these building footprints from Microsoft in Sydney,
>> where we have high resolution aerial imagery with usually pretty good
>> positional accuracy and orthorectification.
>>
>> The MS buildings aren't as good as hand tracing would be, but better than
>> some of the worst mapping done through the HOT tasking manager building
>> tracing projects.
>>
>> Compared to the DCS aerial imagery the alignment (rotation) is usually a
>> fair bit off, but "good enough" basic footprints, the shape is just okay,
>> not great, not terrible.
>>
> ___
> Talk-au mailing list
> Talk-au@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-au
>
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Talk-au Digest, Vol 160, Issue 14

2020-10-20 Thread Warren

Hi John
I use  JOSM.  Any file format that I can bring in as a layer would be 
fine.   I can then select, copy and paste the tracings into an active 
layer for upload, checking as I go.  Certainly faster than tracing by hand.
I am not sure when JOSM get chocked by file size, but say Perth or the 
South West of WA may be enough of a reduction.

Thanks

On 21/10/2020 9:20 am, John Bryant wrote:
Hi Warren, I could probably help with this. What would be a good size 
for a chunk? What would be a useful format?


On Wed, 21 Oct 2020 at 07:21, Warren > wrote:


Hi

I am in the eastern suburbs of Perth where minimal buildings have
been
traced.  I would be happy to check trace data in my area, lets
face it
hand tracing is not much fun  and very time consuming.  I think some
inaccuracies are acceptable, they can be modified as they become
apparent.
The data at
https://github.com/microsoft/AustraliaBuildingFootprints
 is
much too large for me to handle.
Is  someone more skillful than me able to break this data set into
bite
sized chunks?


___
Talk-au mailing list
Talk-au@openstreetmap.org 
https://lists.openstreetmap.org/listinfo/talk-au




___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Andrew Harvey
Also buildings which are touching like a shop fronts are just one polygon,
but I don't think you'd ever be able to do this too reliably without a
survey anyway. Even if there is a small gap between buildings eg a garage
and residence sometimes it will join them into a single polygon.

What I feel it does do well is actually identify buildings, I haven't found
any false positives, only a few false negatives where a building is hidden
under dense tree coverage.

On Wed, 21 Oct 2020 at 12:52, Andrew Harvey 
wrote:

> On Tue, 20 Oct 2020 at 21:20, Simon Poole  wrote:
>
>> Just as a comment: there is nothing so time consuming as fixing badly
>> mapped buildings (essentially drawing them from scratch is nearly always
>> faster), I would only import building outlines that are at a quality level
>> that you would not want to change them except if the building itself has
>> been modified.
>>
>
> I took a look at these building footprints from Microsoft in Sydney, where
> we have high resolution aerial imagery with usually pretty good positional
> accuracy and orthorectification.
>
> The MS buildings aren't as good as hand tracing would be, but better than
> some of the worst mapping done through the HOT tasking manager building
> tracing projects.
>
> Compared to the DCS aerial imagery the alignment (rotation) is usually a
> fair bit off, but "good enough" basic footprints, the shape is just okay,
> not great, not terrible.
>
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Andrew Harvey
On Tue, 20 Oct 2020 at 21:20, Simon Poole  wrote:

> Just as a comment: there is nothing so time consuming as fixing badly
> mapped buildings (essentially drawing them from scratch is nearly always
> faster), I would only import building outlines that are at a quality level
> that you would not want to change them except if the building itself has
> been modified.
>

I took a look at these building footprints from Microsoft in Sydney, where
we have high resolution aerial imagery with usually pretty good positional
accuracy and orthorectification.

The MS buildings aren't as good as hand tracing would be, but better than
some of the worst mapping done through the HOT tasking manager building
tracing projects.

Compared to the DCS aerial imagery the alignment (rotation) is usually a
fair bit off, but "good enough" basic footprints, the shape is just okay,
not great, not terrible.
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Talk-au Digest, Vol 160, Issue 14

2020-10-20 Thread John Bryant
Hi Warren, I could probably help with this. What would be a good size for a
chunk? What would be a useful format?

On Wed, 21 Oct 2020 at 07:21, Warren  wrote:

> Hi
>
> I am in the eastern suburbs of Perth where minimal buildings have been
> traced.  I would be happy to check trace data in my area, lets face it
> hand tracing is not much fun  and very time consuming.  I think some
> inaccuracies are acceptable, they can be modified as they become apparent.
> The data at https://github.com/microsoft/AustraliaBuildingFootprints is
> much too large for me to handle.
> Is  someone more skillful than me able to break this data set into bite
> sized chunks?
>
>
> ___
> Talk-au mailing list
> Talk-au@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-au
>
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Talk-au Digest, Vol 160, Issue 14

2020-10-20 Thread Warren

Hi

I am in the eastern suburbs of Perth where minimal buildings have been 
traced.  I would be happy to check trace data in my area, lets face it 
hand tracing is not much fun  and very time consuming.  I think some 
inaccuracies are acceptable, they can be modified as they become apparent.
The data at https://github.com/microsoft/AustraliaBuildingFootprints is 
much too large for me to handle.
Is  someone more skillful than me able to break this data set into bite 
sized chunks?



___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Simon Poole
Just as a comment: there is nothing so time consuming as fixing badly 
mapped buildings (essentially drawing them from scratch is nearly always 
faster), I would only import building outlines that are at a quality 
level that you would not want to change them except if the building 
itself has been modified.


Simon

PS: lives in an area which had 10s of 1000s of buildings mapped already 
in 2009 from mushy misaligned yahoo imagery.


Am 20.10.2020 um 11:12 schrieb Andrew Harvey:
On Tue, 20 Oct 2020 at 18:39, Daniel O'Connor 
mailto:daniel.ocon...@gmail.com>> wrote:


and after a bit of digging came across this ODBL data:
https://github.com/microsoft/AustraliaBuildingFootprints


As fun as hand tracing data is; I'd be pretty keen to swap to
small scale imports of this, suburb by suburb in my area.

https://wiki.openstreetmap.org/wiki/Microsoft_Building_Footprint_Data

details the data in more general terms, and links to a few import
proposals.

Anyone feel like exploring and validating the quality of the data
in their areas?


Thanks for posting, I didn't realise they had released something for 
Australia. It says "The data will also be made available in Facebook 
RapiD." but I couldn't see buildings come through only roads. RapiD 
would probably be one of the easier options for bringing into OSM.


Areas which are already well mapped building wise best to leave what's 
there, and area where people are actively mapping with local knowledge 
and survey's best to leave what's there, but for general areas with 
almost no buildings mapped, I think it makes sense to manually import 
this to provide a good base level of coverage.


___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


OpenPGP_0x4721711092E282EA.asc
Description: application/pgp-keys


OpenPGP_signature
Description: OpenPGP digital signature
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Andrew Harvey
On Tue, 20 Oct 2020 at 18:39, Daniel O'Connor 
wrote:

> and after a bit of digging came across this ODBL data:
> https://github.com/microsoft/AustraliaBuildingFootprints
>
> As fun as hand tracing data is; I'd be pretty keen to swap to small scale
> imports of this, suburb by suburb in my area.
>
> https://wiki.openstreetmap.org/wiki/Microsoft_Building_Footprint_Data
> details the data in more general terms, and links to a few import proposals.
>
> Anyone feel like exploring and validating the quality of the data in their
> areas?
>

Thanks for posting, I didn't realise they had released something for
Australia. It says "The data will also be made available in Facebook
RapiD." but I couldn't see buildings come through only roads. RapiD would
probably be one of the easier options for bringing into OSM.

Areas which are already well mapped building wise best to leave what's
there, and area where people are actively mapping with local knowledge and
survey's best to leave what's there, but for general areas with almost no
buildings mapped, I think it makes sense to manually import this to provide
a good base level of coverage.
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Microsoft Australian building footprints

2020-10-20 Thread Andrew Harvey
On Tue, 20 Oct 2020 at 18:39, Daniel O'Connor 
wrote:

> Hi folks,
> I was wondering how mapbox had coverage of areas that hadn't been traced
> into OSM; and after a bit of digging came across this ODBL data:
> https://github.com/microsoft/AustraliaBuildingFootprints
>

Not relevant to OSM here, but just wanted to point out that Mapbox's
buildings in Australia are from PSMA not Microsoft ->
https://docs.mapbox.com/vector-tiles/reference/mapbox-streets-v8/
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au


[talk-au] Microsoft Australian building footprints

2020-10-20 Thread Daniel O'Connor
Hi folks,
I was wondering how mapbox had coverage of areas that hadn't been traced
into OSM; and after a bit of digging came across this ODBL data:
https://github.com/microsoft/AustraliaBuildingFootprints

As fun as hand tracing data is; I'd be pretty keen to swap to small scale
imports of this, suburb by suburb in my area.

https://wiki.openstreetmap.org/wiki/Microsoft_Building_Footprint_Data
details the data in more general terms, and links to a few import proposals.

Anyone feel like exploring and validating the quality of the data in their
areas?
___
Talk-au mailing list
Talk-au@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-au