Re: [OSM-Talk-ZA] That Shouldn't Be Possible™ coverage extended to South Africa

2013-12-24 Thread Robert Scott
On Tuesday 24 December 2013, Gerhardus Geldenhuis wrote:
> Hi Robert,
> That looks really cool.
> We recently got a very generous ongoing data donation from a tracking
> company that is providing us with anonymised tracking data for all vehicles
> they provide tracking for. I think a combination of the data and your tool
> would make for something really useful. The problem is the volume... any
> thoughts on how that could be handled? Maybe something similar to
> http://maproulette.org that can automate the detection of problems,
> visualise it and leave it as something to fix/investigate for users.

Yes, I've wanted to make it a far more automatic and less hands-on system, but 
the thing holding that up is the fact that the results from the tool are often 
a little tricky to interpret. GPS traces are generally full of many types of 
"noise" from GPSs that are having a bit of a fit to people getting out of the 
car and going for a small walk with the GPS.

In the future I would like to develop a more automatic way of flagging 
"problems" on a trace as little yellow notes asking "is the speed limit right 
here?", "were you going the wrong way down a one-way street here?", "is this 
really a no-right-turn here?". But I haven't got to the point of figuring that 
out yet.

Once I had such a thing in place it would probably allow for a more automatic 
approach, but as of now this is the best we have.


robert.

___
Talk-ZA mailing list
Talk-ZA@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-za


Re: [OSM-Talk-ZA] That Shouldn't Be Possible™ coverage extended to South Africa

2013-12-24 Thread Gerhardus Geldenhuis
Hi Robert,
That looks really cool.
We recently got a very generous ongoing data donation from a tracking
company that is providing us with anonymised tracking data for all vehicles
they provide tracking for. I think a combination of the data and your tool
would make for something really useful. The problem is the volume... any
thoughts on how that could be handled? Maybe something similar to
http://maproulette.org that can automate the detection of problems,
visualise it and leave it as something to fix/investigate for users.

Greetings from Iceland, no South Africa. Wait the UK actually. ;-)

Regards


On 22 December 2013 18:27, Robert Scott  wrote:

> Hello Iceland,
>
> My GPS trace analyzer, That Shouldn't Be Possible™ has recently extended
> its reach beyond the british isles & benelux to cover South Africa too.
>
> Its purpose? To accept a GPS trace of a drive/cycle you've gone for,
> analyze the journey against the routable OSM database and, if appropriate,
> say "That Shouldn't Be Possible". Used like this, it can find quite a lot
> of routing problems or road segments missing from the database.
>
> It can also be used to take the hard work out of checking the OSM database
> against your trace after a long journey by flagging up sections that don't
> quite agree with OSM.
>
> Not quite sure whether you've got that complex junction interlinked and
> tagged right? Have a gps trace or two that traverses it? That Shouldn't Be
> Possible might be able to help you.
>
> An example analysis result can be seen here[1]: I've left a nice great big
> error (visible as a spike in the plot) in the middle of it as an example
> where the trace seems to traverse what's marked as a path.
>
> I've written a lot more about it on the wiki[2], so I'm not going to
> duplicate all that blather here. It's still in what I would call a
> prototype stage, but it works surprisingly well for "motorcar" traces (less
> so for bicycle so far, though I would say that's largely the fault of the
> current state of OSRM's bicycle profile). Even so[3] I encourage mappers to
> try it out[4] on one of their traces.
>
> Merry xmas everybody.
>
>
> robert.
>
> [1] http://ris.dev.openstreetmap.org/tsbp-proto/779918/2/2/
> [2] http://wiki.openstreetmap.org/wiki/That_Shouldnt_Be_Possible
> [3] Especially so - I need testers.
> [4] http://ris.dev.openstreetmap.org/tsbp-proto/
>
> ___
> Talk-ZA mailing list
> Talk-ZA@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-za
>



-- 
Gerhardus Geldenhuis
___
Talk-ZA mailing list
Talk-ZA@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-za


Re: [OSM-Talk-ZA] That Shouldn't Be Possible™ coverage extended to South Africa

2013-12-22 Thread Robert Scott
On Sunday 22 December 2013, Robert Scott wrote:
> Hello Iceland,

Damn it.


robert.

___
Talk-ZA mailing list
Talk-ZA@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-za


Re: [OSM-Talk-ZA] That Shouldn't Be Possible™ coverage extended to South Africa

2013-12-22 Thread David Richfield
That sounds really awesome! I'll have to try it out.

Sent from a mobile device.
On 22 Dec 2013 19:25, "Robert Scott"  wrote:

> Hello Iceland,
>
> My GPS trace analyzer, That Shouldn't Be Possible™ has recently extended
> its reach beyond the british isles & benelux to cover South Africa too.
>
> Its purpose? To accept a GPS trace of a drive/cycle you've gone for,
> analyze the journey against the routable OSM database and, if appropriate,
> say "That Shouldn't Be Possible". Used like this, it can find quite a lot
> of routing problems or road segments missing from the database.
>
> It can also be used to take the hard work out of checking the OSM database
> against your trace after a long journey by flagging up sections that don't
> quite agree with OSM.
>
> Not quite sure whether you've got that complex junction interlinked and
> tagged right? Have a gps trace or two that traverses it? That Shouldn't Be
> Possible might be able to help you.
>
> An example analysis result can be seen here[1]: I've left a nice great big
> error (visible as a spike in the plot) in the middle of it as an example
> where the trace seems to traverse what's marked as a path.
>
> I've written a lot more about it on the wiki[2], so I'm not going to
> duplicate all that blather here. It's still in what I would call a
> prototype stage, but it works surprisingly well for "motorcar" traces (less
> so for bicycle so far, though I would say that's largely the fault of the
> current state of OSRM's bicycle profile). Even so[3] I encourage mappers to
> try it out[4] on one of their traces.
>
> Merry xmas everybody.
>
>
> robert.
>
> [1] http://ris.dev.openstreetmap.org/tsbp-proto/779918/2/2/
> [2] http://wiki.openstreetmap.org/wiki/That_Shouldnt_Be_Possible
> [3] Especially so - I need testers.
> [4] http://ris.dev.openstreetmap.org/tsbp-proto/
>
> ___
> Talk-ZA mailing list
> Talk-ZA@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-za
>
___
Talk-ZA mailing list
Talk-ZA@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-za


[OSM-Talk-ZA] That Shouldn't Be Possible™ coverage extended to South Africa

2013-12-22 Thread Robert Scott
Hello Iceland,

My GPS trace analyzer, That Shouldn't Be Possible™ has recently extended its 
reach beyond the british isles & benelux to cover South Africa too.

Its purpose? To accept a GPS trace of a drive/cycle you've gone for, analyze 
the journey against the routable OSM database and, if appropriate, say "That 
Shouldn't Be Possible". Used like this, it can find quite a lot of routing 
problems or road segments missing from the database.

It can also be used to take the hard work out of checking the OSM database 
against your trace after a long journey by flagging up sections that don't 
quite agree with OSM.

Not quite sure whether you've got that complex junction interlinked and tagged 
right? Have a gps trace or two that traverses it? That Shouldn't Be Possible 
might be able to help you.

An example analysis result can be seen here[1]: I've left a nice great big 
error (visible as a spike in the plot) in the middle of it as an example where 
the trace seems to traverse what's marked as a path.

I've written a lot more about it on the wiki[2], so I'm not going to duplicate 
all that blather here. It's still in what I would call a prototype stage, but 
it works surprisingly well for "motorcar" traces (less so for bicycle so far, 
though I would say that's largely the fault of the current state of OSRM's 
bicycle profile). Even so[3] I encourage mappers to try it out[4] on one of 
their traces.

Merry xmas everybody.


robert.

[1] http://ris.dev.openstreetmap.org/tsbp-proto/779918/2/2/
[2] http://wiki.openstreetmap.org/wiki/That_Shouldnt_Be_Possible
[3] Especially so - I need testers.
[4] http://ris.dev.openstreetmap.org/tsbp-proto/

___
Talk-ZA mailing list
Talk-ZA@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-za