Raster Vision Backend Plugin Example

This repository holds an example of a new backend for Raster Vision, implemented as a plugin.

Setup and Requirements

Requirements

You'll need docker (preferably version 18 or above) installed.

Setup

The Docker container in this repo inherits from raster-vision-cpu and raster-vision-gpu, so you'll need to have those built (using ./docker/build in the raster-vision repo).

To build the image, run the following command:

> scripts/build

This will add some of this repo's code to the standard raster-vision image.

Running the console

Whenever the instructions say to "run the console", it means to spin up an image and drop into a Bash shell by doing this:

> scripts/console

This will mount the following directories:

${HOME}/.aws -> /root/.aws
${HOME}/.rastervision -> /root/.rastervision
notebooks -> /opt/notebooks
plugin -> /opt/src/plugin
spacenet -> /opt/src/spacenet
${RASTER_VISION_DATA_DIR} -> /opt/data
${RASTER_VISION_REPO_DIR}/rastervision -> /opt/src/rastervision

You will need to set the RASTER_VISION_REPO_DIR environment variable so it points to your local raster-vision repo. This will allow you to make changes to the raster-vision source code and have them immediately available in the running container. You will also need to set RASTER_VISION_DATA_DIR to the directory containing your data.

Running chip classification on Vegas

To drive development of a new chip classification backend, we will use the Vegas example in spacenet/vegas.py. It allows running all three tasks on the SpaceNet Vegas example. It is modified from the version in the examples repo to use a backend plugin (in plugin/noop_backend.py) when doing classification. See the build_backend method to see how this backend is utilized in the experiment configuration. You can run this example using something like the following. At the moment, it doesn't generate any real output because the backend is a noop. You may want to read about setting up the Vegas example here.

export ROOT_URI=/opt/data/lf-dev/rv-output/vegas-spacenet
rastervision -p plugin \
    run local -e spacenet.vegas \
    -a test True \
    -a use_remote_data False \
    -a root_uri ${ROOT_URI} \
    -a target buildings \
    -a task_type chip_classification

The -p plugin flag is to use a Raster Vision profile called plugin. This profile points to the location of the plugin module. You can make such a profile by creating a file at ~/.rastervision/plugin containing something like: