isabella232 / geo-ml-docker

Docker Image for Machine Learning with Geospatial Data in Python

geo-ml-docker

This docker is built from developmentseed/geolambda image which includes libraries and their dependencies for geospatial processing, such as numpy and gdal.

In addition to the libraries provided by geolambda, geo-ml-docker includes libraries for machine learning such as tensorflow and keras. See all included libraries in requirements.

Docker Commands

On your local machine

• Build the image

docker build -t geo-ml-docker .

• Basic use case: Run a shell in a new container

docker run -it geo-ml-docker /bin/bash

• Advanced use case

Run a jupyter notebook in a new container, expose the port, and mount a volume. Note - this mounts a volume in the current directory so it should be run in the directory where any required files are maintained.

docker run -it -v ${PWD}:/code -p 8888:8888 geo-ml-docker /bin/bash -c "jupyter notebook --debug --ip=* --port=8888 --no-browser --allow-root"

On an AWS EC2 machine

• Docker installation on the EC2 machine

Follow the AWS's instruction to install Docker on the EC2 machine before you go ahead.

• Build the image

ssh to the EC2 with your pem, and build the docker image.

docker build -t geo-ml-docker .

• Run the jupyter notebook EC2

docker run -v ${PWD}:/code -it -p 8888:8888 geo-ml-docker /bin/bash -c "jupyter notebook --debug --ip=* --port=8888 --no-browser --allow-root"

• Log into the jupyter notebook from your local machine

In another terminal ssh to cloud machine.

ssh -i Your_AWS.pem -L 8888:localhost:8888 [email protected]

• Open your jupyter notebook on the browser

When you run the notebook the first time, copy and pasted something like http://127.0.0.1::8888/?token=xxxxxx to your browser to open the notebook.