Availability of Pre-diffed datasets

[ramSeraph]

I am guessing the current datasets released were already diffed with OSM road data or were the existing OSM roads never attempted for detection?

If all the roads were attempted for detection and this is a dataset published after taking out the roads that were close matches to OSM roads, is it possible to get a pre-diffed dataset under the same MIT license.

The OSM ODBL data licensing might be problematic for a lot of use cases.. for example a government trying to enhance its data and release the data under a less restrictive open license.

[ramSeraph]

As an example for India, were all the OSM roads used in the training and test set or only a subset of them used.. if only a subset was used, then releasing the roads data which weren't part of the training or test set shouldn't cause a licensing problem.. if that was a worry.

[ramSeraph]

There is an actual usecase.. if you want details, i can give them

[zlavergne]

Hi @ramSeraph, thanks for reaching out! Currently, our system doesn't allow us to create a road dataset that isn't conflated with OSM. There would likely be certain aspects of the resulting data that would make it hard to work with (like road classifications, connections, and noisy artifacts).

It would be great to learn more about your use case, however, so we can better understand how our data might be helpful. Feel free to describe your use case in this issue (or link a site/doc if that's easier).

[ramSeraph]

Hey @zlavergne thanks for getting back to me.. this is related to the PMGSY Geosadak Rural road dataset that was released by the Indian Government earlier this year - https://geosadak-pmgsy.nic.in/opendata/

Basically MoRD( Ministry of Rural Development ) has released a lot of missing Indian rural roads as open unrestricted data under OGDL, so as to enrich the maps of OSM and other map providers in India with the data.

But they do want to setup a feedback loop where they can check their data with other data sources like OSM and add to their dataset. But I suspect they can't pull in and release OSM data because of ODBL. So, I am wondering if the data you have under MIT LIcense can be used for that.

This data could also possibly be used to verify the FB roads themselves in some places where the roads were charted through ground surveys( though currently it is not known which part of the data is from ground surveys )

Related OSM wiki page with details - https://wiki.openstreetmap.org/wiki/India/PMGSY_rural_connectivty_data_import

Related github repo with the data - https://github.com/datameet/pmgsy-geosadak

Full Presentation - https://youtu.be/3tI7XIZzhSM?t=9246
Call for feedback from the same video - https://www.youtube.com/watch?v=3tI7XIZzhSM&t=10660s

[ramSeraph]

And also maybe I can deal with the messy parts of the pre-diffed data.. If there is documentation of what the messy parts are.