Contents

Explore Imagery with terracotta

For quick exploration of imagery in the browser, we can use terracotta. This document captures how to serve a folder of images and explore it interactively.

Build terracotta

A containerised version of terracotta can be built using the Dockerfile:

docker built -t terracotta -f ./Dockerfile.terracotta .

This is a relatively lightweight image based off the Alpine version of miniconda3 (v.4.8.2-alpine).

Serving

The approach follows a client/server model so, first, we need to spin up a server that makes the imagery available. For that, we first launch the server container:

docker run --rm \
           -ti \
           -p 5000:5000 \
           -v /media/dani/DataStore/GHS-composite-S2/:/data \
           terracotta sh

This needs to be run off the same machine where the imagery is stored.

Once inside the container, launch the server with:

/opt/conda/bin/terracotta serve \
                          --allow-all-ips \
                          -r /path/to/data/folder/S2_percentile_UTM_{tile}-{scene1}-{scene2}_osgb.tif

This should open a Flask server that will listen to port 5000 for requests.

Exploring

Once the imagery is being served, we can launch a web app through a separate (local or remote) container:

docker run --rm \
           -ti \
           -p 5005:5005 \
           -v /path/to/data/folder/:/data \
           --network host \
           terracotta sh

Note that although the command is similar to that we use for the server, here we set --network host so the client can be reached from the guest’s browser.

Inside the container, then launch the app:

/opt/conda/bin/terracotta connect --port 5005 localhost:5000

Then point your browser to localhost:5005.

Optimising

The above process “works”, but if the imagery is too heavy, it’ll choke and will struggle to be responsive enough for exploration. If that is the case, consider pre-optimising the images. Following the terracotta tutorial, we can generate optimised versions by:

/opt/conda/bin/terracotta optimize-rasters -o /target/folder \
                                           --reproject \
                                           /path/to/data/*.tif

Once this finishes, you can serve as above, with the modification of the data path.