Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Map label placer with simulated annealing.
Tree: 5ccc9d0a70

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.


Dymo is a placement script for map labels, isolated from the purpose-built code in GeoIQ’s Acetate. Dymo resolves positions for densely-packed point labels, and results in layouts make your maps look like they’ve been touched by a cartographer.

Dymo uses Simulated Annealing to derive an acceptable global label set, described in Steven Wouderberg’s October 2007 presentation at Utrecht University. See an animation of the process over time in these two videos of U.S. and European place names:

Using Dymo is a script that converts lists of cities with included font information to GeoJSON point and label files suitable for use in geographic rendering.

Place U.S. city labels at zoom 6 for two minutes:

python -z 6 --minutes 2 --labels-file labels.json --places-file points.json data/US-z6.csv.gz

Place U.S. city labels at zoom 5 over a 10000-iteration 10.0 - 0.01 temperature range:

python -z 5 --steps 10000 --max-temp 10 --min-temp 0.01 -l labels.json -p points.json data/US-z5.csv

Both examples will result in a pair of GeoJSON files, labels.json and points.json. The first will contain rectangular label areas, the second will contain center points of places successfully positioned by Dymo. Because labels will collide in different ways depending on map scale, labels must be placed separately for each zoom level:

U.S. Labels

A trick for determining the right temperature range for a given data input is to run once for a short time, e.g. --minutes 0.1, noting the temperature ranges reported in the output. You can then vary the number of steps up and down to arrive at a good-looking label set without waiting for the first “Exploring temperature landscape” step each time.

Data Included

Look in data/ for a list of zoom-by-zoom city locations, organized by continent and selected by Justin O’Bierne. Data is distributed separately from code, see downloads for a link.

To prepare your own city lists or modify fonts and font sizes in input lists, Use to apply population-specific font choices to an in-bound list:

python --font 0 fonts/Arial.ttf 10 --font 1000000 fonts/Arial.ttf 12 --zoom 5 data/North-America-all.txt.gz data/North-America-z5.txt

Sample Output

These sample images were created during the development of Acetate, and show the results of town placement.

Central Japan:

Central Japan

Western Europe:

Western Europe

Northen Appalachians:

Northen Appalachians


Copyright 2010-2011 Michal Migurski and GeoIQ, offered under the BSD license. Uses Richard J. Wagner’s Python annealing library.

We’re not affiliated with Dymo Corporation.

Something went wrong with that request. Please try again.