There are several methods available for installing solaris:
Installing with only pip use at your own risk!
Regardless of installation method, you’ll need Python version 3.6 or greater.
More details on installing Python can be found
here. Additionally, if you
plan to use the SpaceNet dataset with
solaris (it features prominently in
many of the tutorials), you’ll need a free Amazon Web Services account
and the AWS CLI installed
If you’re just going to work with your own data, you can skip these steps.
Installing from GitHub using a
conda environment and
If you wish to install a bleeding-edge version of
solaris that isn’t available
on conda-forge yet, you can install from GitHub. You’ll need
anaconda for this installation as well.
From a terminal, run:
git clone https://github.com/cosmiq/solaris.git cd solaris git checkout [branch_name] # for example, git checkout dev for bleeding-edge
If you have access to a GPU where you’re installing
solaris, use the following:
conda env create -f environment-gpu.yml
If you don’t have access to a GPU:
conda env create -f environment.yml
Finally, run the last two lines (for installs both with or without GPU):
conda activate solaris pip install .
The above installation will create a new conda environment called
containing your desired version of solaris and all of its dependencies.
Installing with only
Use this method at your own risk!
If you have already installed the dependencies with underlying binaries that
don’t install well with
rtree), you can easily
pip install the rest:
pip install solaris
Note that this will raise an error if you don’t already have