[HOT] Is HOT planning to validate this 4 year old project?

Fri Dec 17 14:07:49 UTC 2021

There are a number of problems with validation as it stands.

First the faster you get feedback to the mapper the better it is.  I did 
some experiments on this.  Catch them within a hour or two and you 
prevent problems plus you get a lot more out of the mappers.  After a 
week the feedback isn't so valuable.  They may have changed their method 
of working so any repetition is basically a turn off.  After a month 
forget it.

Second it takes time to open up a tile, check it over.  For duplicate 
buildings when a building has been uploaded twice you need to run a 
script to detect them.  You need to run JOSM validation to check 
crossing highways etc.  I don't think Mapswipe can begin to detect these 
sort of problems.

What can be done is to grab a chunk of the map and take a look at the 
problems.  For example Angola  has a thousand duplicate buildings at 
this point in time.  The scripts will identify them and you can take 
action.  Whether that should be a changeset comment leaving time for a 
response or simply deleting one of the duplicate buildings as many 
validators would do is a matter of debate.  Have you taken this into 
account with mapswipe by the way?  Can you compare different background 
imagery?

One of the biggest problems I see with buildings is the area around the 
buildings tagged as a building.  Place=plot might be appropriate.  These 
can be picked out with an overpass query based on building size.

So there are two parts of validation, the first is changing habits but 
you have to get in early to do this.  The second is map clean up and 
often this can be done quicker by using tools over a wider area.

Cleaning up an area works well since new mappers tend to map copy the 
tags that are already in use in the map.  It's also a lot cheaper in 
validator time than opening  up individual tiles.

By the way if the American Red Cross really wants to make a difference 
to building quality then either train in JOSM with the buildings_tool 
plugin or add one to iD.  For a line of buildings it takes two mouse 
clicks in the JOSM buildings_tool plug in against I think five in iD to 
map a building.  For buildings not in line it takes three mouse clicks. 
That still gives you roughly twice as many buildings for the same mapper 
time and they are correctly tagged.

Cheerio John

Benjamin Herfort wrote on 12/17/2021 7:16 AM:
>
> Hey,
>
> just adding a few thoughs on validation and OSM data quality 
> implications. It's true that there are probably still to many projects 
> in the Tasking Manager that have not been validated. On the other 
> hand, it is also important to acknowledge that over the past years 
> validation efforts have been improved. (Probably not to the level 
> everyone would like to see it, but at least it goes into the right 
> direction.) And there are many volunteer validators that work on that 
> every day.
>
> The MapSwipe community and HeiGIT are currently working on a project 
> funded by American Red Cross, which will allow MapSwipe users to 
> validate OSM building footprints. I've briefly talked about this at 
> this years HOT summit. It's clear that this will probably not help us 
> to prevent low quality buildings edits in the first place. But maybe 
> this new MapSwipe project type can help us to validate Tasking Manager 
> projects quicker or at least provide a new way to find out which 
> buildings need a validation / correction. (During the HOT summit talks 
> there were also many other discussions around validation and how it 
> can be supported by more analytics.)
>
> For some background information: You can check Tasking Manager stats 
> on a HeiGIT's Humanitarian OSM Stats page: 
> https://humstats.heigit.org/statistics.html If you scroll down to the 
> "Recent Performance" section you get an overview on the share of 
> Tasking Manager activity that went into mapping and validation. 
> Ideally the dark blue bar would be even bigger than 50%, which would 
> mean that more tasks get validated than are newly mapped. If you like 
> you can check this also for a few organisations such as Crowd2Map 
> (Tanzania Development Trust) or Doctors Without Borders (MSF), which 
> are doing relatively well in regard to validation.
>
> Finally, we can also check which projects have been mapped, but not 
> validated and are already rather old. E.g. download this file: 
> https://humstats.heigit.org/api/export/all/project_progress_sessions_centroid.geojson 
> and filter in QGIS:
>
> "progress_mapped" > 0.9 AND "progress_validated" < 0.1 AND 
> "start_activity" <= '2017-01-01'
>
> This will give you a list of around 200 projects which are created 
> before 2017, mapped to minimum 90% and validated to less than 10%. I'm 
> pretty sure that there are other people here in the community that 
> have similar approaches to identify these projects. Maybe there is 
> something in the Tasking Manager already, I'm not aware of as well.
>
> Have a nice day,
>
> Benni
>
> PS: I've sent that message yesterday already containing some figures, 
> but it seems that it didn't get through properly.
>
> disclaimer: I'm working at HeiGIT and are also heavily involved with 
> MapSwipe.
>
>
> On 15.12.21 15:56, John Whelan wrote:
>> I've run over the area and deleted a couple of hundred duplicate 
>> buildings.  The problem with validating buildings is it takes longer 
>> to validate them than map them using JOSM's buildings_tool plugin.  
>> Given teh imagery is often fussy ad the buildings do not generally 
>> line up with all the imagery they are a pain in the neck to do.  So 
>> not formally validated but has been cleaned up.  My person view is 
>> the NGO are after building counts to get an estimated population so a 
>> simple rectangle would suffice for their purposes.
>>
>> One other thing I noticed was courtyards in the center of a 
>> building.  Fairly common but how  do you may them and more 
>> importantly explain to a new mapper how to map them?
>>
>> Cheerio John
>>
>> Mateusz Konieczny via HOT wrote on 12/15/2021 3:24 AM:
>>> https://tasks.hotosm.org/projects/3431
>>>
>>> I encountered more garbage data added on instructions of HOT.
>>>
>>> This is not a problem by itself, as with new mappers it will 
>>> occasionally happens.
>>>
>>> The problem is that after four years it is still not validated.
>>>
>>> I would strongly recommend that HOT would use more of its resources
>>> on validating already made edits rather on adding more things.
>>>
>>> Is there some plan to actually validate edits organized by HOT?
>>>
>>>
>>> _______________________________________________
>>> HOT mailing list
>>> HOT at openstreetmap.org
>>> https://lists.openstreetmap.org/listinfo/hot
>>
>> -- 
>> Sent from Postbox <https://www.postbox-inc.com>
>>
>> _______________________________________________
>> HOT mailing list
>> HOT at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>
>
> _______________________________________________
> HOT mailing list
> HOT at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot

-- 
Sent from Postbox <https://www.postbox-inc.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20211217/c0db1e99/attachment.htm>