Project X - University of Alberta

Setup

First install Anaconda, a Python distribution / package manager.

The first command here creates a new conda environment named projectx and installs our base packages.

# create the conda environment
$ conda env create -n projectx -f env/base.yml

# activate the conda environment
$ conda activate projectx

# to deactivate the conda environment
$ conda deactivate

# update the conda environment with an environment yml file
$ conda env update -n projectx -f <path to .yml file>

# to remove the conda environment
$ conda env remove -n projectx

If you need a python package that isn't in the base.yml file, add it there and update the environment.

CPU / GPU specific libraries

Some libraries (mainly deep learning libraries like PyTorch and Tensorflow) must be installed differently if you have a CUDA-capable GPU, these should be installed separately from the base.yml file.

PyTorch Installation

# CPU only
$ conda env update -n projectx -f env/torch-cpu.yml

# GPU
$ conda env update -n projectx -f env/torch-gpu.yml

Google Earth Engine Authentication

In order to use the Earth Engine API you first need to authenticate with the earthengine CLI. Make sure you're in the projectx conda environment and run earthengine authenticate. Follow the steps there (using the UAIS google account). It will save an auth token to your machine letting you use the API.

Code Style

Before you push any Python code, format it using black. If you make a pull request with unformatted code a check will fail on the PR.

$ black .

Git Workflow

Don't push directly to master, instead work on a separate branch and when it's ready you can open a pull request. If the changes are non-trivial somebody should review and approve the pull request before you merge it.

Project Structure

.
├── .github
|   └── workflows   github actions
├── src             source code
├── notebooks       jupyter notebooks
├── data            data (large data should be hosted in the google drive and gitignored)
├── models          trained models (again, large files in google drive, not on the repo)
├── papers          LaTeX files, bibliographies, etc.
└── env             conda environment files

Try to structure things logically within each directory.