I like the idea for a project of the week using OS OpenData StreetView, but
would suggest that before we add lots of new roads we work hard to get roads
which are already in OSM properly named. Firstly it is improving data which is
already there, secondly it using a second, independent, data source. Not a
patch on ground survey, but at least it means that editors of the data have to
engage with the data sources and their discrepancies. I would not be happy at
OSM becoming a largely a subset of Ordnance Survey data without more thought
(but also see below).
As for the status of noname roads, I have named perhaps 2000 or so in West
London and Merseyside in the past few weeks. There are still substantial parts
of the South East and North West with many unnamed roads. I have not estimated
the number, but its still in the thousands. Unfortunately the noname map layer
on the website has not been updated (along with other Cloudmade maps), so I'd
suggest using beta.letuffe.org which has a noname overlay (link is to Wigan
area).
It is important not to forget that a mass import of the VectorMap District
roads named from Locator will become possible within the next six months. I'm
sure several people are looking at a) how to accurately name the VMD roads from
Locator ; and b) how to find only those roads which are not already in OSM
(e.g., by using the techniques of the French CORINE project). Once viable
technical solutions to these issues are available we will be able to import ALL
the missing roads SHOULD we wish to. Manual tracing of StreetView data should
be considered in this context.
Personally, I don't think mass imports of VectorMap District road data should
be
contemplated, at least for 6 months or so, for all the usual reasons (Pottery,
Imports and the Community). However, availability outwith the planet database
of those roads in
VectorMap District and not in OSM could be used to enhance downstream
applications, such as Garmin extracts, and specific map renders. In other words
we should be able to generate GB road-complete products without risking some of
the known effects on community building of armchair mapping.
I think there is plenty of scope to think of other 'added-value' projects with
the StreetView data, these are some off the top of my head:
* Getting all schools in to coincide with publication of league tables
(its another data source to cross-check)
* Mapping all professional football grounds (see for instance Blundell
Park)
* Ditto for other sports (e.g., crags used for climbing, horse
racecourses, ...).
* Mapping landuse=residential for areas without streets (shapes can be
used as a guide to poorly mapped areas)
* Get all churches tagged with man_made=tower or man_made=spire if
applicable so that we can do OSGB like renders
* Get all bridges tagged and marked for major waterways. Bridges across
large rivers are surprisingly poorly mapped. It ought to be possible to
identify these and make our existing data better.
* Replace larger expanses of water mapped from NPE or Yahoo with OSSV
or OS VDM.I hope these thoughts are not too controversial. I must add that I am
not a zealot for the "no import" cause, but I do recognise that there is a
reasonable case for it.
Regards,
Jerry Clough
________________________________
From: Kai Krueger <[email protected]>
To: talk-gb <[email protected]>
Sent: Sun, 6 June, 2010 12:07:33
Subject: [Talk-GB] UK Project of the week - trace a village off of OSSV?
Hello everyone,
I would like to suggest as a sort of "Project of the week" for the UK
for people to pick a random town or village somewhere in the UK that so
far has poor coverage and trace it's roads from OS OpenData StreetView.
Despite the various claims over the years that the UK road will be "road
complete" by the "end of the year", the UK is still a far distance off
of that target. I have heard the numbers that so far we have on the
order of 50% of named roads (people who are working on OS - OSM
comparisons please correct me if I am wrong). Which is by no means a
small feat of achieving, but also not as high as one would like it to be.
So let us try and accelerate this a bit by everyone picking a small
random town or village somewhere in the UK and trace the roads from
StreetView. It probably only takes about 10 - 20 minutes for a small
village and even a small town isn't too bad to do (if the weather is bad
and you can't go out). So with the help of OS data, we can get a big
step closer to where we would like to be and use it as a basis to
continue to improve beyond the quality of OS data or any other
commercial map provider.
(If you are convinced already, then no need to read the rest of the email)
I know that many people are opposed to "armchair mapping" or imports
(and btw I am not proposing a full scale import here, but manual tracing
instead) and so I'd like to counter some of the arguments most likely
going to be brought up against this sort of non local tracing:
1) OS data might have mistakes, be outdated and generally not as good as
what OSM aims for: Yes, no doubt OS has errors and can be outdated in
many places by a couple of years ( I have found more than enough of
those myself). Furthermore, all of the OS products released lack many of
the properties we are interested in like one way roads, turn and other
restrictions, POIs, foot and cycle ways and all the other things that
make OSM data such a rich and valuable dataset. So yes, the OS data will
clearly not replace any of the "traditional" OSM surveying techniques or
be the end of things. But it can be a great basis to build upon.
As a comparison, have a look (assuming you have a timecapsal ;-)) at
what the data of e.g. central London looked like in 2007. It already had
surprisingly many roads, but hardly any POIs or other properties that we
aim for now. Most of that came later in many iterations of improvement.
A single pass of "OSM" surveying is not any better than the OS data per
se. Also given that the errors introduced by tracing OS data are exactly
the same type of errors introduced by manual "OSM" surveying, i.e.
misspellings in roads, missing roads, outdated roads, ... We need to
have the tools to deal with this kind of maintenance anyway. It is the
iterations that make OSM data what it is, not the "first pass ground
survey".
Creating a blanket base layer from OS data allows us to much better
focus on the aspects that do distinguish us from every other map data
provider with having to "waste" as little as possible resources on the
"stuff everyone else has" too.
2) large scale imports and tracing hinders community growth: This
perhaps is the more important of the two arguments, as indeed what
distinguishes us from everyone else is the community and without the
community and its constant iterations and improvements, OSM data will
"bit rot" just as much as all other data. However I don't think there is
any clear evidence either way of what non local mapping does to
communities and it remains hotly debated. The negative effects claimed
are usually of the form a) The area looks complete, there is nothing
more to do, so why bother. Or, it isn't as much fun to add a POI than a
whole new village on a blank canvas. b) I put in all this effort into
mapping an area and along comes an import and steam rollers all this
into a mess, I am leaving. c) imports introduce a new class of bugs,
e.g. duplicate nodes or broken connectivity that OSM mappers wouldn't so
we don't have the tools to deal with these sort of errors correctly.
b) and c) are specific to imports and thus manual tracing shouldn't
suffer the same issues. a) may be the case, but it is clearly a case
that we need to be able to deal with anyway, as more and more areas
become "complete" by "them selves". And looking at the better mapped
areas, like Germany or some of the UK cities, I don't think there is any
evidence that you can't attract new comers into already mapped areas. It
is potentially also offset by all those people who simple want to use
the data for something like embed a map into their blog or use OSM data
on their Garmin, their phone, their game, their ... and will fix the odd
bug they discover while doing so, but can't really as it simply isn't
complete enough yet.
Other examples of remote mapping have also been fairly successful. The
most obvious one was Haiti. It's initial phase was entirely arm chair
mapping and had no community at all. Only later followed by on the
ground surveying. Never the less it is generally considered a success
and has gained OSM many new mappers.
The other example is mapping during holidays. Lets say I go and visit a
mostly unmapped island in Scotland. I'll be able to survey a few roads,
add the odd POI make a few mistakes and miss many details. I will also
never return to that place again to fix up any bugs I might have
introduced. Should I not have mapped during the holidays as I wasn't a
"local mapper" or part of the "local community"? If I do it in a foreign
country, I might not even no the local laws.
So again, we as a community as a whole need to be able to deal with
these sorts of issues that also arise from armchair mapping and it is a
great test for our ability to create appropriate quality assurance tools.
Anyway, far too much rambling from my side already, so I better stop now
again. I just felt like countering some of the general negativity
towards armchair mapping and imports.
Kai
_______________________________________________
Talk-GB mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/talk-gb
_______________________________________________
Talk-GB mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/talk-gb
