Now as I know how to extract data for any region. A new challenge is how to refine data as it does have lots of line data which are not roads in my case in Barmer rajasthan we have sand dunes which has shadow property with contrast in google earth. I am thinking to minimize this data with accuracy in three steps 1. First to clean data upto village level Get list of all villages of India with Lat Log address (most probably middle of village) 2. create an buffer of 500 Mtr or 1 km or 2 km as per population index 3. select Facebook ai data in that buffer region only and delete rest of lines 4. merge that data with OSM line data ( only lines which are connected to an existing osm lines ) in other words if a line is not connected with any other line(road ) with OSM data with the help of query tool it will be deleted
I believe this can be a method to refine data. If you agree or have better idea kindly guide with some instruction video or page as number of villages in india are more then 15 lak and manually this is next to impossible If we can publish that data on this group this will be ahead all the data source available even with bhuvan wms rj_roads doesnt have that much data. Note - I have a shape file data that distinguish between urban and rural areas of India but accuracy is upto 85 % (area larger then actual area) On Friday, May 1, 2020 at 7:27:01 AM UTC+5:30, Nikhil VJ wrote: > > Hi, I learned from GIS stackexchange > <https://gis.stackexchange.com/questions/266601/clip-specific-layer-by-bounding-box-with-ogr2ogr> > > that "-spat" may be better than "-clipsrc" as it doesn't bother to cut > features that extend beyond our rectangle. Which I'm fine with, and it > saves a lot on processing time. > > So this command worked better for me: > ogr2ogr -spat 70.1 18.2 75.5 20.6 > output.gpkg IN_PK_mapwithai_road_data.gpkg > > PFA a screenshot : I loaded OSM road data (highway=*) in white and this > data (let's call it FB-AI) in orange over a satellite view. > > My observation: The FB-AI data has excluded existing roads that are > already there on OSM it seems, and is marking possible roads that don't > existi in OSM yet. And It seems to be doing a pretty good job of detecting > unmapped roads. > > Caveats: Inevitable consequence of trying to catch em all: False > positives. It's marking any scratch it finds and many are just dirt tracks > in the fields or even dried canals that may or may not be getting used as > general thoroughfare. While local tracks used by tractors to go to fields > are fine for mapping as dirt tracks (pls don't mark them as residential > roads or unclassified), there is a greater priority to mark roads that can > be used by regular vehicles without damaging those vehicles; roads than can > be used in times of emergencies, for delivery of essentials and the likes. > And you wouldn't want to send an ambulance trying to get to a landlslide > affected area into an impasse while thinking they were taking a shortcut as > shown on map. > > For this reason, IMHO the current approach of not declaring this data as > roads straightaway but rather feeding it in as suggestions and letting > human mappers decide (see links below), is better. That being said, there > can be purposes other than precise mapping, such as using this data as > proxy for indicating other things, or aggregating over areas to compare > between areas, or using this to map irrigation canals (:D), or using it to > decide where to focus the next mapping efforts. This data can be useful and > fill in (do mention the caveats in your summary!) for areas where OSM > mapping hasn't had proper coverage yet. > > > I'm sharing source links: > Github page where all countries' data download links are posted: > > https://github.com/facebookmicrosites/Open-Mapping-At-Facebook/wiki/Available-Countries > Ah, even there they're mentioning "Country exports contain only the AI > predicted roads that are missing from OpenStreetMap." > > The RapID editor that brings this data as pink suggestions into OSM's > online editor for mapping: > https://mapwith.ai/rapid > > The main website where there's explanation etc: https://mapwith.ai/ > > PS: Satellite imagery brought in to QGIS by copying over tile urls from > http://leaflet-extras.github.io/leaflet-providers/preview/ under "XYZ > Tiles" in left pane where we find the OpenStreetMap layer. > > > Regards > Nikhil VJ > https://nikhilvj.co.in > > > -- Datameet is a community of Data Science enthusiasts in India. Know more about us by visiting http://datameet.org --- You received this message because you are subscribed to the Google Groups "datameet" group. To unsubscribe from this group and stop receiving emails from it, send an email to datameet+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/datameet/b0638440-7c05-4f0a-a41e-6cf3279d9c86%40googlegroups.com.
