Data pipeline for machine learning with OpenStreetMap
JavaScript HTML Makefile
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
classes Added road-buildings-rail-natural class Aug 22, 2017
disaster Add script to tag osm buildings w/ damage assessment Jan 3, 2017
lib Merge branch 'master' of Sep 11, 2017
.dockerignore Docker Aug 13, 2016
.eslintrc Initial commit Apr 7, 2016
.gitignore Update docker.env ignore to include postfixes to catch backup files Aug 19, 2017
Dockerfile Made container run as non-privileged user Aug 19, 2017
LICENSE Create LICENSE Aug 25, 2017
Makefile Use data dir for temporary files, not workdir Aug 19, 2017 Don't recommend the docker detached flag Nov 2, 2017
docker-compose.yml Mount local dir with docker-compose, but take node_modules from built… Aug 19, 2017
download-images Aerial imagery should be jpg, labels png Aug 16, 2017
label-counts Add bbox filter option to sample May 22, 2016
label-stats Add bbox filter option to sample May 22, 2016
list-image-pairs Update list-image-pairs to use jpg for imagery Aug 16, 2017
package.json Add helper script for displaying tile lists as geojson Feb 13, 2017
palette-to-grayscale Add class statistics Apr 7, 2016
preview.html Update preview.html Nov 13, 2017
rasterize-labels Aerial imagery should be jpg, labels png Aug 16, 2017
remove-bad-images Add white threshold for image deletion Jan 3, 2017
run-tile-reduce Clean tile reduce script May 23, 2016
sample Guard against train size > total tiles Jul 25, 2016
sample-to-geojson Add sample-to-geojson Nov 30, 2016


A pipeline to simplify building a set of training data for aerial-imagery- and OpenStreetMap- based machine learning. The idea is to use OSM QA Tiles to generate "ground truth" images where each color represents some category derived from OSM features. Being map tiles, it's then pretty easy to match these up with the desired input imagery.

Quick Start

Pre-built docker image

The easiest way to use this is via the developmentseed/skynet-data docker image:

First, create a docker.env file with the contents including your MapboxAccessToken:


Then run:

docker run -v /path/to/output/dir:/workdir/data --env-file docker.env developmentseed/skynet-data download-osm-tiles

docker run -v /path/to/output/dir:/workdir/data --env-file docker.env developmentseed/skynet-data

The first line downloads the OSM QA tiles to /path/to/output/dir/osm/planet.mbtiles. If you've already got that file on your machine, you can skip this.

The second builds a training set using the default options (Roads features from OSM QA tiles, images from Mapbox Satellite). To change the data sources, training set size and other options, add the relevant environment variables to the docker.env file , one per line.

Local docker image

You can also create the docker images yourself using docker-compose. Similarly to the quick-start above, make sure your docker.env file has your MapboxAccessToken and any other environment variables you want to set. Then run:

docker-compose build

to build your local docker image, and

docker-compose run data download-osm-tiles
docker-compose run data 

to download the OSM QA tiles, and run the data collection as specified in docker.env. By default the collected data will be saved into the data directory, but you can overide it by using -v /path/to/output/dir:/workdir/data after docker-compose run data similar to the pre-built instructions above.


The make commands below work off the following variables (with defaults as listed):

# location of image files
IMAGE_TILES ?= "tilejson+$(MapboxAccessToken)"
# which osm-qa tiles extract to download; e.g. united_states_of_america
# location of data tiles to use for rendering labels; defaults to osm-qa tiles extract specified by QA_TILES
DATA_TILES ?= mbtiles://./data/osm/$(QA_TILES).mbtiles
# filter to this bbox
BBOX ?= '-180,-85,180,85'
# number of images (tiles) to sample
# define label classes output
# Filter out tiles whose ratio of labeled to unlabeled pixels is less than or
# equal to the given ratio.  Useful for excluding images that are all background, for example.
# set this to a zoom higher than the data tiles' max zoom to get overzoomed label images

You can override any of these parameters in your docker.env and make a full training set using the instructions above.



  • Install NodeJS v4.6.2
  • Install tippecanoe
  • Install GNU Parallel
  • Install shuf
  • Clone this repo and run npm install. (Note that this includes a node-mapnik install, which sometimes has trouble building in bleeding-edge versions of node.)

Sample available tiles

make data/sample.txt

This just does a simple random sample of the available tiles in the given mbtiles set, using tippecanoe-enumerate. For more intelligent filtering, consider using tippecanoe-decode to examine (geojson) contents of each tile.


Build label images: make data/labels/color or make data/labels/grayscale. Uses the CLASSES json file to set up the rendering of OSM data to images that represent per-pixel category labels. See classes/water-roads-buildings.json for an example. Rendering is with mapnik; see the docs for more on filter syntax.


Download aerial images from a tiled source: make data/images

Heads up: the default, Mapbox Satellite, will need you to set the MapboxAccessToken variable, and will cost you map views!


Preview the generated data by opening up preview.html?accessToken=<mapbox access token>&prefix=/path/to/data in a local web server.