[Imports] Import building footprints Lopa & Lingo

Wed Jul 22 10:30:19 UTC 2020

Hi Mateusz,
I guess it is not that bad as you think it might be. I’ve just taken quickly two screenshots of the data:
This is how most of the data is: https://www.dropbox.com/s/vgqpv6lsrcxa5hy/Capture_1.PNG?dl=0
But some of the data looks like this – mostly in the more densely built areas: https://www.dropbox.com/s/a8pfjn1ai26f0id/Capture_2.PNG?dl=0
So even if there are sometimes a bit weird geometries, they still do represent reality. You can clearly see the shape they try to make, it is just not always perfectly done (they should have used JOSM for this 😉).
On local survey, that will actually been done (if security of course allows). The data will be used during a malaria spraying campaign, something similar like the one described in this presentation: https://f1000research.com/slides/9-453 Especially for this type of work the material of the housing is important to know what insecticide you need to spray.
And I probably should have thought a bit more how to write about the geometries. We do care, hence this discussion and also the proposal for a manual verification of the buildings and not a bulk upload.
Thanks, Jorieke

Jorieke VYNCKE
Missing Maps Coordinator & OCA GIS Focal Point
Manson Unit - MSF UK

[MSF_dual_English_CMYK]
MSF UK, 10 Furnival Street, London, EC4A 1AB
www.msf.org.uk<http://www.msf.org.uk/>  ||  +44 (0)20 7404 6600 || UK Charity Reg. No. 10265888
Don't print if you don't need

From: Mateusz Konieczny <matkoniecz at tutanota.com>
Sent: 22 July 2020 10:53
To: Jorieke Vyncke <jorieke.vyncke at london.msf.org>
Cc: 'imports at openstreetmap.org' <imports at openstreetmap.org>
Subject: Re: [Imports] Import building footprints Lopa & Lingo

Jul 22, 2020, 11:23 by jorieke.vyncke at london.msf.org<mailto:jorieke.vyncke at london.msf.org>:
Further I am a bit confused, if we would have been able to digitize the data from the imagery we would have done so. We have done this already in the places surrounding the data we have available here for the Lopa & Lingo area. See here: https://www.openstreetmap.org/way/820347035#map=14/1.7331/30.4503 However now we have this good quality dataset that is created on imagery of January 2020, which is several years more recent than we have available for usage on OpenStreetMap and we would not be able to use that on OpenStreetMap because we cannot verify this data with imagery on OpenStreetMap?
Obviously, you can verify it also with other methods (such as local survey).

It may be OK to blindly import datasets (and conflate them with what is mapped already),
but only high quality ones without broken/invalid/suspicious geometries.

If datasets contains some broken geometries then it is no longer OK to assume that
shapes included in it are valid, even if given shape is not obviously broken.
So trying to convince you again why in my opinion it is a good quality dataset:
Some geometries are completely broken and clearly not representing reality, right?

  *   There is a very low risk in features ‘not representing anything real’. The exact geometry is mostly not that important for the usage of the data for us
What you mean by "very low risk"? 0.1% of geometries is bogus? Lest than 100 geometries in
entire dataset? 1% of geometries is broken?

Remark that you are not caring about exact geometry is also worrying to me.
What you mean by that? There is some simplification even in standard mapping but I have also
seen  people mapping all building as squares because exact geometry was not important for them.

  *   this in contrary to the indication of the existance of the building
Maybe you would prefer to map landuse areas?

  *   and the material it is made off.
Are you sure that this can be reliably imported?
So yes, there might be a slight issue with the geometries of a small number of buildings, but this is thus because of written reasons in my opinion something we can fix by the manual treatment of the data.
What you mean by "small number of buildings"? How you plan to fix this given that you are unable
to do it doing import?

So what do others think?

Would the situation change if myself and the few volunteers working on uploading the data could have access to the initial imagery? I think that might be maybe possible…  Of course if there are experienced mappers here who would like to help out with the import, please let me know, you’re always welcome to join the effort!

Btw, thanks to challenge me. I hope we can come to a good solution here together?

Thanks,
Jorieke

From: Mateusz Konieczny via Imports <imports at openstreetmap.org<mailto:imports at openstreetmap.org>>
Sent: 21 July 2020 13:04
Cc: imports at openstreetmap.org<mailto:imports at openstreetmap.org>
Subject: Re: [Imports] Import building footprints Lopa & Lingo

As I understand, the plan is to load the buildings in manageable sets from the tasking manager,
and person doing verification will skip solely clearly bizarre geometries and assume that everything
else is correct.

I think that given presence of nonsense geometries means that seemingly OK ones cannot
be assumed to be valid ones, and only verified ones should be imported.

Jul 21, 2020, 11:53 by winfixit at gmail.com<mailto:winfixit at gmail.com>:
I get the impression Mateusz thinks the import will consist of an automatic dump of the data, whereas the actual plan is to load the buildings in manageable sets from the tasking manager and then a person does the verification, maybe by using the todo plugin in JOSM?

Polyglot

On Tue, Jul 21, 2020 at 10:25 AM Jorieke Vyncke <jorieke.vyncke at london.msf.org<mailto:jorieke.vyncke at london.msf.org>> wrote:
Hi Mateusz,

Thanks for your feedback!
There are really not that many of this type of buildings in the dataset, most of the buildings are nicely squared or circulized. By the manual verification of the data, we’ll of course compare and adapt the building data with the latest imagery that we have available on OpenStreetMap – the Maxar Premium Imagery. We’ll correct with help of that imagery and of course in all cases nicely square corners of the weird shapes. There will really not be a ‘soul-crushing cleanup’ needed after this import.

Would you rather propose to leave these buildings out? We could do that as well, but I would prefer not to, since this means we will have an incomplete dataset in the area. I’d rather choose for correcting the shapes with help of available imagery.

Best wishes,

Jorieke

From: Mateusz Konieczny via Imports <imports at openstreetmap.org<mailto:imports at openstreetmap.org>>
Sent: 20 July 2020 18:20
Cc: 'imports at openstreetmap.org<mailto:imports at openstreetmap.org>' <imports at openstreetmap.org<mailto:imports at openstreetmap.org>>
Subject: Re: [Imports] Import building footprints Lopa & Lingo

Jul 20, 2020, 18:18 by jorieke.vyncke at london.msf.org<mailto:jorieke.vyncke at london.msf.org>:
Q: Some buildings have strange trapezoidal shapes, how are you planning to correct them without accessing the imagery?
There are not that many of these. Since we are not able to check back with the original imagery, we’ll have to trust the dataset we want to import for this and will upload the data. Of course by looking at the data manually we would be able to spot areas with real issues with this and then we can still decide to not upload that data.
So clearly suspicious shapes such as this trapeizodal ones will be skipped and
everything else without blatant mistakes will be imported without verification?

That seems to be a bad idea and I am opposed to such import. It will just burden
any community of mappers that may appear with soul-crushing cleanup.
_______________________________________________
Imports mailing list
Imports at openstreetmap.org<mailto:Imports at openstreetmap.org>
https://lists.openstreetmap.org/listinfo/imports

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20200722/25269d41/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.jpg
Type: image/jpeg
Size: 2312 bytes
Desc: image002.jpg
URL: <http://lists.openstreetmap.org/pipermail/imports/attachments/20200722/25269d41/attachment-0001.jpg>