ScenicStroll is a web app that finds scenic routes between two locations based on the density of geotagged photos from Flickr. The scenery model uses a kernel density estimate to create a probability map of photo locations. Optimal routes are then found by a graph search on OpenStreetMap road data where the graph edge weights are determined from a combination of distance and scenery score, where the relative weight of the two metrics is controlled by the user.

The backend uses PostgreSQL with PostGIS extensions for geospatial queries. To build the database, first download OpenStreetMap XML data, e.g. from Mapzen. Next, create a new database and enable PostGIS extensions:

psql> CREATE DATABASE scenicstroll;
psql> \connect scenicstroll
psql> CREATE EXTENSION postgis;

Finally, generate the routing database by running, e.g.

python parse_osm.py san-francisco-bay_california.osm postgres:///scenicstroll
psql -c prep_routes.sql

Computation of scenery scores is handled by notebooks/scenery_score.ipynb.

To install the dependencies for the web server:

pip install $(cat requirements.txt)

Then start the server with

python app.py