[Imports] Uganda Bureau of Statistics educational facilities import

[Rafael Avila Coya]

Hi all:

On 30/12/15 15:26, Christoph Hormann wrote:
> On Wednesday 30 December 2015, Jotam wrote:
>>
>> I wanted to reply to Christoph's objection about schools being hard
>> to identify on aerial imagery. In fact, it is pretty easy to identify
>> Ugandan schools on satellite pictures as almost all school grounds
>> are set up to a very particular pattern and shape. This may sound
>> improbable to Western ears. But Ugandans like to simply copy a once
>> tried and tested concept. That's why almost all school compounds look
>> alike.
> 
> You are right, this does sound unlikely w.r.t. a volume of 24k schools 
> all over Uganda - not from a narrow minded western viewpoint but based 
> on basic understanding of schooling practice all over the world and 
> observation of currently mapped schools in Uganda:
> 
> http://mc.bbbike.org/mc/?lon=32.493016&lat=0.845208&zoom=17&num=3&mt0=mapbox-satellite&mt1=mapnik&mt2=google-satellite
> http://mc.bbbike.org/mc/?lon=32.477664&lat=0.849601&zoom=17&num=3&mt0=mapbox-satellite&mt1=mapnik&mt2=google-satellite
> http://mc.bbbike.org/mc/?lon=32.469376&lat=0.831407&zoom=17&num=3&mt0=mapbox-satellite&mt1=mapnik&mt2=google-satellite
> 
> The fact that you can recognize some schools in Uganda by common 
> architectural elements (which by the way also applies to other parts of 
> the world) does not change anything here - this might be helpful for 
> remote mapping to identify some schools, it is not of any use for 
> verifying separately acquired data.
> 
> Note i am not saying the data is bad and should not be imported but 
> since it is not possible to verify it based on available imagery it 
> would be necessary to see and demonstrate if the data is reliable 
> before making plans for an import.
> 
> Currently the import documentation does not even contain the basic 
> information that can be assessed without on-the-ground knowledge, like 
> how many schools are currently mapped in OSM within the area of the 
> import, how many are both in OSM and in the new data and how many are 
> only in one of them and not in the other.  IMO this is a basic 
> requirement for a useful import discussion.  On-the-ground assessment 
> of the data in a few sample areas would be good to have in addition of 
> course.

There are 1,133 schools as of today mapped in Uganda. Of them, 686 are
nodes and 447 ways. There isn't any edu facility mapped as relation in
Uganda.

Of those objects, by inspection with JOSM, at least 63 of the nodes
aren't in the UBOS set, and 31 of the ways aren't in the UBOS set. So
those wouldn't have to be merged.

>From the rest, inspecting 50 random OSM edu facilities against the UBOS
set, 41 of them (82%) are the same as UBOS, but UBOS correcting the OSM
node and adding more info. The 9 rest are most of them ok. Only 1 (2%)
can't be sure if ok or not.

I hope this, together with other data shown already in the wikis and
other emails in this thread, clarifies the quality of this dataset.

Cheers,

Rafael.


>