[OSM-talk] Why the HOT obsession with low quality buildings in Africa ?

john whelan jwhelan0112 at gmail.com
Mon Jul 2 12:48:18 UTC 2018

Low quality building mapping is pretty general in Africa, I mentioned
Malawi still has some 4,805 duplicate buildings in the HOT mailing list
very recently which as a percentage of the buildings mapped is probably
fairly low but is still a concern.  The problem is one of data quality, is
OSM reliable?  and if I want to promote a commercial alternative the data
quality of buildings in Africa is a weak point of OSM.  Can you trust the
buildings mapped in OSM?

The mappers seem to be mappers for a day or possibly three and having
spoken to one or two training is seen as a waste of time they just want to
map.  I get the impression that mapping buildings is seen as a way of
engaging people and bringing something to their attention as much as
mapping accurate buildings.

Perhaps a better way would be to map the village outline and tag them with
the number of buildings on a date?

Cheerio John

On 2 July 2018 at 07:26, Robert Banick <rbanick at gmail.com> wrote:

> Many humanitarian groups use buildings as a rough proxy for population
> (density), or to ensure every household is covered during a vaccination
> campaign, or simply to navigate. Likely they use them for a combination of
> the three. As Phil says, it’s best to read the specific task.
> As a side note, it’s helpful to be more specific than the entire continent
> of Africa, which is a very, very large and diverse place. If you can note
> individual problematic countries, as Frederic does, it helps us to identify
> sources of error or verify there was a legitimate humanitarian or
> community-building reason behind any fall in quality.
> On Mon, Jul 2, 2018 at 4:16 PM Warin <61sundowner at gmail.com> wrote:
>> On 02/07/18 18:52, Frederik Ramm wrote:
>> > Hi,
>> >
>> > On 02.07.2018 10:24, Jean-Marc Liotier wrote:
>> >> churning out buildings like demented stonemasons trying to reach their
>> weekly quota
>> >> of gamified task-managing !
>> > I recently stumbled upon
>> >
>> > https://www.openstreetmap.org/#map=15/-6.8958/39.1623
>> >
>> > (Tanzania) and had a similar thought. The buildings there are at least
>> > square and largely match aerial imagery, but this, too, looked like
>> > supercharged one-trick-pony image tracing combined with very little on
>> > the ground knowledge (e.g. quite a few roads and tracks clearly visible
>> > on the imagery are not traced, and from someone local you'd expect the
>> > occasional POI or label).
>> >
>> > Someone must have buildings very high on their priority list (don't even
>> > know if HOT are involved but it certainly doesn't look like local
>> mapping).
>> >
>> > It will be interesting to learn why buildings are so important. Or are
>> > they just the lowest-hanging image tracing fruit, or just easier to
>> count?
>> >
>> I have been mapping a few buildings lately - mainly to add addresses to.
>> Past mappers have placed a few POI ... but they tend not to be too
>> precise - e.g. between buildings or on the footpath.
>> Once the building outline is there then you see the discrepancy. And any
>> further additions of POI can be guided by the building outlines.
>> I hope 'my' buildings are a little better that what is described above,
>> some of that depends on the imagery,
>> some in the pride of workmanship and some on the fatigue of the mapper.
>> Certainly any HOT manager who rewards the number of things done should be
>> alert to the quality reduction that such motivation brings.
>> One of the good things about adding addresses .. you notice things like
>> the road name is wrong.
>> _______________________________________________
>> talk mailing list
>> talk at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/talk
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20180702/b177cf5a/attachment.html>

More information about the talk mailing list