[OSM-talk] Licence of Facebook's derived road datasets? ODbL?

Joseph Eisenberg joseph.eisenberg at gmail.com
Fri Nov 15 12:41:49 UTC 2019


I don't know anything about the legal aspects, but logically I don't
understand why it is being claimed that facebook releasing a list of
possible roads, which specifically excludes all roads in
Openstreetmap, is not using Openstreetmap data.

I understand that training their algorithim ("AI") on Openstreetmap
data is fine.

But the data that they have released is specifically a list of
features which look like roads on aerial imagery, excluding those
which are already included in the Openstreetmap database, right?

How would anyone use the dataset without also combining it with
Openstreetmap data?

Now if facebook wants to release a dataset of "all things that look
like roads in aerial imagery according to our algorithm (which BTW was
trained on OSM), that's fine.

But they have already gone one step further and then added all
`highway=` features in Openstreetmap to the dataset - in this case by
subtracting those features which are already a very closely aligned to
a `highway` way.

- Joseph Eisenberg

On 11/15/19, Martin Koppenhoefer <dieterdreist at gmail.com> wrote:
> Am Fr., 15. Nov. 2019 um 12:41 Uhr schrieb Christoph Hormann
> <osm at imagico.de
>>:
>
>> Because the basis of most comments made does not seem to be the desire
>> to neutrally assess the situation Rory presents here and its
>> implications.
>> What it seems instead happens here is that people look at the situation
>> and develop a spontaneous reaction in terms of "should this be possible
>> or not" and then specifically search for ways to argue in support of
>> this opinion.
>
>
>
> I am exempting myself from this, because I would not like Facebook to be
> able to use OSM data and not follow the license, but I believe they can in
> this case. ;-)
>
>
>
>> From an engineering perspective the idea that adding OSM data can create
>> a derivative database but subtracting OSM data cannot does not hold up
>> of course.  I can create a polygon data set of the Earth surface (a
>> simple rectangle in EPSG:4326) and subtract an OSM derived data set of
>> the Earth land masses from that to get a data set of the oceans.
>> According to the hypothesis this would not be subject to the ODbL.
>>
>
>
> You are generalizing in a way that is not suitable. What was stated was
> that there must be OSM data (in original or derived form) in the data to
> make the license kick in. In the case presented by Rory, IMHO there isn't
> OSM data in their dataset. It will not be possible to deduct any kind of
> OSM data from their dataset. In your example, you clearly have derived OSM
> data in your new dataset, otherwise it wouldn't be possible to get back to
> the original data (or part of it). The question is not "addition or
> subtraction", but whether there is data from OSM in the data.
>
> Cheers
> Martin
>



More information about the talk mailing list