[Talk-us] Comparing Tiger 2017 dataset with OSM in a automatedway.

Ian Dees ian.dees at gmail.com
Thu Oct 12 13:52:13 UTC 2017


Thanks for posting this Badita. When I look through this in JOSM I see a
very small portion of TIGER roads that are missing from OSM that should
actually be there. For example, there are some clumps of roads in
subdivisions or apartment complexes like this:

[image: Inline image 1]

The vast majority of roads seem to be correctly missing from OSM. For
example, this section of roads that are placed over the top of a baseball
stadium:

[image: Inline image 2]

This is the same thing I've seen with my TIGER 2017 delta map I made. After
a decade of hard work, the OSM community has far surpassed the TIGER
dataset in terms of quality. I would suggest that it is more useful to run
CYGNUS comparisons with OSM over data directly from local governments
(there are several examples in https://github.com/osmlab/centerlines)
rather than the possibly sub-par data from TIGER. The local government data
will be more likely to have recently-added or recently-removed roads.

-Ian

On Thu, Oct 12, 2017 at 8:19 AM, Badita Florin <baditaflorin at gmail.com>
wrote:

> I started processing, State by State, the Tiger 2017 dataset, because i
> get less errors from overpass this way then if i process in bulk of 10
> states.
>
> The first state is Alabama.It took around 5 hours to complete,and it shows
> over 20.000 possible ways that are not added in OSM.
> The total length of the ways is 5042 km, or 3133 miles.
> The xml size is 36Mb, and can be found in this folder.
> https://drive.google.com/drive/folders/0B7aOUf0DFRnLU3hUWFNu
> S05JS2c?usp=sharing
>
> There are 5 35x35 miles tiles missing from the data.
>
> Be mindful that this is the raw output, and even if Tiger 2017 is better
> then older versions, there are still a lot of bad data in Tiger.
>
> I expect more then 50% of the data to not be good for mapping into OSM.
> Also, you can filter this data-set and keep just the road with the names,
> that could be more important roads compared to the ones that don`t have a
> name.
>
> @joe, regarding the dataset for the twin cities, if you can create/provide
> a translation file for the shapefile, i can run it and post the results in
> the folder.
>
> Some examples here
>
> https://github.com/ToeBee/ogr2osm-translations
>
>
>
>
>
>
>
>
>
> On Wed, Oct 11, 2017 at 6:46 PM, Ian Dees <ian.dees at gmail.com> wrote:
>
>> Yes, it'd be great to do this. I started a project to track open road
>> centerlines like this on GitHub here: https://github.com/osmla
>> b/centerlines
>>
>> In theory, we could download that data and then do the same process using
>> the presumably higher quality road data.
>>
>> On Wed, Oct 11, 2017 at 10:42 AM, Joe Sapletal <joe.sapletal at charter.net>
>> wrote:
>>
>>> This is really cool.  Can I suggest for the Twin Cities metro area
>>> someone doing something similar with the Metro Regional Centerlines
>>> Collaborative Local Centerlines (MRCC)?  I know that Dakota County hasn’t
>>> submitted centerlines to Tiger in a couple of years, but will be for the
>>> next update.  Not sure about the other counties though.  There very well
>>> may be areas that the MRCC will be a better source than the Tiger data.
>>>
>>>
>>>
>>> https://gisdata.mn.gov/dataset/us-mn-state-metrogis-trans-mr
>>> cc-centerlines
>>>
>>>
>>>
>>> Joe
>>>
>>>
>>>
>>> *From: *Ian Dees <ian.dees at gmail.com>
>>> *Sent: *Wednesday, October 11, 2017 10:25 AM
>>> *To: *Badita Florin <baditaflorin at gmail.com>
>>> *Cc: *talk-us at openstreetmap.org Openstreetmap
>>> <talk-us at openstreetmap.org>
>>> *Subject: *Re: [Talk-us] Comparing Tiger 2017 dataset with OSM in a
>>> automatedway.
>>>
>>>
>>>
>>> It would be interesting to see what differences CYGNUS would turn up.
>>> What would the output of CYGNUS be?
>>>
>>>
>>>
>>> I put together the TIGER 2017 layer that's in the editors right now.
>>> I'll work on writing up how I did it later today.
>>>
>>>
>>>
>>> Basically: I used tiger-tiles (https://github.com/iandees/tiger-tiles)
>>> to generate a vector tiles database with expanded road names from TIGER
>>> 2017. Then I downloaded an osm-qa-tiles (https://osmlab.github.io/osm-
>>> qa-tiles/) file for the United States and ran osmlint's tigerDelta (
>>> https://github.com/osmlab/osmlint/tree/master/validators/tigerDelta) to
>>> find the segments that had different geometry. The output was then ran
>>> through Tippecanoe to generate a vector tileset and posted to Mapbox as the
>>> low zoom red features.
>>>
>>>
>>>
>>> On Wed, Oct 11, 2017 at 4:03 AM, Badita Florin <baditaflorin at gmail.com>
>>> wrote:
>>>
>>> Hi, i wanted to ask if there will be interest around comparing TIGER
>>> 2017 with what we have in OSM, using CYGNUS, in a automated way.
>>> http://www.openstreetmap.org/user/mvexel/diary/36746
>>>
>>>
>>>
>>> On top of cygnus, i have developed shgp2cygnus, were you can place any
>>> shapefile, any size, you provide a translation file, and, in the end, you
>>> get a list with all the ways that are in the TIGER dataset, but not in OSM.
>>>
>>> This would be something useful for the community ?
>>>
>>>
>>>
>>> I don`t know if somebody is already doing something similar, or what is
>>> the status ? Were can i read more ?
>>>
>>> I knoiw that the TIGER 2017 Overlay in JOSM shows in red the roads that
>>> are not in OSM, but are in TIger 2017.
>>>
>>> But I don`t know were to read more, and if the data is accessible to
>>> download directly, not just show as a WMS Layer.
>>>
>>>
>>>
>>> It will take around 7-14 days to process all USA”
>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20171012/b4448bf4/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 481245 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20171012/b4448bf4/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 641868 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20171012/b4448bf4/attachment-0003.png>


More information about the Talk-us mailing list