Repository for our work on the Understanding Clouds from Satellite Data competition on Kaggle
Your work on this project should follow our Contribution guidelines.
dss_kaggle_clouds/models, obviously includes the models, and dss_kaggle_clouds/features has any auxillary features/helper functions, to clean data to remove the bars from stitched images, or to do image augmentation to increase the size of training data.
This uses pipenv to manage a vitual environment, and we're limiting ourself to 3.6.2 to run on Hummingbird. The requirements.txt file is created from the Pipfile and specifies locked dependencies for setup.py, to allow reproducable biulds.
If when you type python -V, you see Python 3.x.x (some version of 3), you can replace python3 with python for the installation instructions.
Install pipenv, or skip this step if you already have pipenv installed.
python3 -m pip install --user pipenv
# or
python -m pip install --user pipenv
# or
pip install --user pipenv
If you can type pipenv in your shell and you it doesn't complain about that not being a command, you can replace python3 -m pipenv with just pipenv, else you can invoke it with python3 -m pipenv as a module
Make sure you're in the directory with the Pipfile, and then:
Install packages:
python3 -m pipenv install --skip-lock
If pipenv complains about you not having a python version 3.6 installed, you can install another version of python with pyenv. (brew install pyenv on mac)
And then you can install and specify the python version like:
pyenv install 3.6.2
python3 -m pipenv --python ~/.pyenv/versions/3.6.2/bin/python3
Enter the virtual environment
python3 -m pipenv shell
Install our code as a module in the virtualenv:
python3 setup.py install
Run the script:
python3 run.py --help
If you ever change any code in the dss_kaggle_clouds module/directory (make sure you're in the virtual environment), run:
python3 setup.py install # to reinstall the package
# and then
python3 run.py data/raw # ... run the script
You can type exit to leave the virtual environment. (and python3 -m pipenv shell to re-enter)