GeoSmell

Experimental code for computing signals about geographic areas.

To identify geographic areas, we use S2 cells. You can choose the desired aggregation level; --level=17 aggregates to cells of about 70×70 meters in size. To visualize the geographic region of an S2 cell on a map, you can use the Sidewalk Labs Region Coverer. Enter the S2 cell ID into the “cells” field and press the circular button.

Currently, we process the following datasets:

• wikicommons: geo-tagged pictures at Wikimedia Commons
• chpopstat: Swiss Population Statistics

Building and running

git clone https://github.com/brawer/geosmell.git ; cd geosmell
go build ; go test
./geosmell --dataset=wikicommons --level=17

Storage

If you pass the access keys to an S3-compatible cloud storage system, the tool will store its output into a cloud storage bucket “geosmell”. Also, if you pass the storage keys, the tool will check what data version is in storage; if the stored data version is still current, the tool will exit early without doing work. This makes it convenient to use geosmell as a daily running cronjob. For --keys, pass the location of a JSON file with the following content:

{
    "S3Host": "your.preferred.cloud.storage.example.com",
    "S3AccessKey": "YourAccessKey",
    "S3SecretKey": "YourSecret"
}

Future work

It would be nice to collect additional signals about geographic areas. Feel free to send pull requests.

License

The code in this repository is Copyright 2019 by Sascha Brawer, licensed under the MIT license. The processed data sets have their own licences.