[HOT] Squared buildings

Sun Apr 17 21:19:16 UTC 2016

Just to recap the problems, we start with iD not displaying the buildings
in the tile, whether this is because the time has lapsed and a second
mapper has started on the tile or iD not showing all the detail I'm not
concerned with the reason simply the fact that to me its not reliable.

Then we have mappers not using mice and not zooming in.

Huts seem routinely to be one mapped correctly then all the next ones
people come across are cut and paste of the first.  This means the size of
many is incorrect.

We have an expectation that if we look at the area of the building we can
estimate the population.

So we end up with a lot of approximately mapped buildings which we then ask
people to square.  When we square we are approximating again which means
the accuracy for building area goes down even further.

Whilst Jo is happy to carefully inspect each building after squaring I
probably don’t have the patience when faced with a large number and I
suspect a fair number of validators feel the same.  Especially when its
faster to go in delete the lot and remap with JOSM building_tool plugin.

I think we can assume that a four sided building will have four sides when
mapped.

I personally think that a squared building looks better but from a
functional point of view we know there is a building there, the aid workers
have a map which shows them the location and if the four sides aren't
perfectly square they will still be able to recognise it.

My personal view is for four sided buildings some sort of image recognition
software as the first pass followed by validation would give us much better
accuracy and probably be faster.

My second choice would be to use something like the building_tool plugin
for JOSM. It would give us much better accuracy and people might even
manage to get the building lined up with the four corners of the image.

We could of course clone Jo but that might be difficult.

Cheerio John

On 17 April 2016 at 14:40, Jo <winfixit at gmail.com> wrote:

> Hi John,
>
> I'm validating tasks with many buildings in it and even though we stressed
> on it for the Mapathon, I still find quite a few of them not being made
> rectangular.
>
> So I started using this search to find all the buildings with 4 nodes:
>
> building inview nodes:4
> Square them all, then search like this:
> building parent modified
>
> So you can add all the buildings which have nodes that moved to the todo
> list. Then you can use ] quickly to review them and see if it still makes
> sense. use 'w' to move their nodes if needed, followed by 'q'. Then ']'
> again to move to the next one. This makes it relatively efficient without
> losing accuracy. It definitely beats ]q]q]q]q]q] :-)
>
> Then search again using:
>
> building inview nodes:5-
>
> to review the ones with more nodes.
>
> building inview nodes:-9
>
> also works to exclude round buildings.
>
> Jo
>
>
>
> 2016-04-15 1:23 GMT+02:00 john whelan <jwhelan0112 at gmail.com>:
>
>> >2. Validation - either invalidate or fix.
>>
>> >
>> Step 1 is the preferable route but if people are working on their own or
>> the turnout makes one on one assistance impossible, then it should be fixed
>> in the validation step.
>>
>> I think less well under half of the mapped tiles in HOT have been
>> validated and of those that have I'd say another 20+% wouldn't meet my
>> personal standards and 50+% wouldn't meet Jo's.  I admit my personal
>> validation standard is aimed more at making sure what is there is
>> reasonably correct according to the project instructions.
>>
>>  So are you suggesting gold standard validation ie JOSM plugin todo list
>> and each building is examined carefully before squaring?
>>
>> Is some form of bulk squaring acceptable?  On the grounds its better than
>> nothing?
>>
>> If the tiles get invalidated who do we expect to come back and fix them?
>> Remember 99% of the "unoffical" maperthon mappers will never return.
>>
>> In the case of projects that have many of these types of buildings which
>> may not be attractive to validate should we just ignore the problem and
>> hope one day someone will gold plate validate the project.  It may even
>> happen.
>>
>> Remember that validation is voluntary and validators can choose which
>> projects to validate on and which to just ignore.
>>
>> I accept some of the big organised groups probably think they have proper
>> training on their organised maperthons and tame validators to map their
>> particular projects so for them the problem doesn't exist but think in
>> terms of HOT generally, think in terms of the maperthons that take place
>> with no experienced mappers.  They exist.
>>
>> I understand it is not an easy question and there are very different view
>> points but I think we need to have the discussion and attempt to reach some
>> sort of consensus of how to get the most out of the limited resources we
>> have rather than have individual validators make their own pragmatic
>> decisions.  One of which is delete them all and remap, its faster.
>>
>> Cheerio John
>>
>>
>>
>>
>>
>> On 14 April 2016 at 18:33, Clifford Snow <clifford at snowandsnow.us> wrote:
>>
>>>
>>> On Thu, Apr 14, 2016 at 3:16 PM, john whelan <jwhelan0112 at gmail.com>
>>> wrote:
>>>
>>>> So your suggestion on how to deal with the existing poorly mapped
>>>> buildings would be?
>>>
>>>
>>> 1. Determine the cause(s) of the poorly mapped buildings. Do we need
>>> more helpers in MM mapathons? The last one I did, we had a number of new
>>> mappers. Those of us helping were stretched just answering questions. Not
>>> being able to spend time going over people work. And yes - we did teach
>>> squaring buildings. We also recommended people bring a mouse to the
>>> session. One of our team brought extra for people to use and I even lent
>>> mine out. Drawing features without a mouse is difficult. We've even
>>> suggested to Red Cross that they have a bag of mice to lend during MM
>>> events.
>>>
>>> 2. Validation - either invalidate or fix.
>>>
>>> Step 1 is the preferable route but if people are working on their own or
>>> the turnout makes one on one assistance impossible, then it should be fixed
>>> in the validation step.
>>>
>>> Best,
>>> Clifford
>>>
>>>
>>> --
>>> @osm_seattle
>>> osm_seattle.snowandsnow.us
>>> OpenStreetMap: Maps with a human touch
>>>
>>
>>
>> _______________________________________________
>> HOT mailing list
>> HOT at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20160417/fe3f2833/attachment.html>