Going beyond BEDMAP2 using a super resolution deep neural network. Also a convenient flat file data repository for high resolution bed elevation datasets around Antarctica.
Launch in Binder (Interactive jupyter notebook/lab environment in the cloud).
Start by cloning this repo-url
git clone <repo-url>
Then I recommend using conda to install the non-python binaries (e.g. GMT, CUDA, etc). The conda virtual environment will also be created with Python and pipenv installed.
cd deepbedmap
conda env create -f environment.yml
Activate the conda environment first.
conda activate deepbedmap
Then set some environment variables before using pipenv to install the necessary python libraries,
otherwise you may encounter some problems (see Common problems below).
You may want to ensure that which pipenv
returns something similar to ~/.conda/envs/deepbedmap/bin/pipenv.
export HDF5_DIR=$CONDA_PREFIX/
export LD_LIBRARY_PATH=$CONDA_PREFIX/lib/
pipenv install --python $CONDA_PREFIX/bin/python --dev
#or just
HDF5_DIR=$CONDA_PREFIX/ LD_LIBRARY_PATH=$CONDA_PREFIX/lib/ pipenv install --python $CONDA_PREFIX/bin/python --dev
Finally, double-check that the libraries have been installed.
pipenv graph
conda env update -f environment.yml
pipenv sync --dev
Note that the .env file stores some environment variables.
So if you run conda activate deepbedmap
followed by some other command and get an ...error while loading shared libraries: libpython3.7m.so.1.0...
,
you may need to run pipenv shell
or do pipenv run <cmd>
to have those environment variables registered properly.
Or just run this first:
export LD_LIBRARY_PATH=$CONDA_PREFIX/lib/
Also, if you get a problem when using pipenv
to install netcdf4, make sure you have done:
export HDF5_DIR=$CONDA_PREFIX/
and then you can try using pipenv install
or pipenv sync
again.
See also this issue for more information.
conda activate deepbedmap
pipenv shell
python -m ipykernel install --user --name deepbedmap #to install conda env properly
jupyter kernelspec list --json #see if kernel is installed
jupyter lab &
The paper is published at The Cryosphere and can be referred to using the following BibTeX code:
@Article{tc-14-3687-2020,
AUTHOR = {Leong, W. J. and Horgan, H. J.},
TITLE = {DeepBedMap: a deep neural network for resolving the bed topography of Antarctica},
JOURNAL = {The Cryosphere},
VOLUME = {14},
YEAR = {2020},
NUMBER = {11},
PAGES = {3687--3705},
URL = {https://tc.copernicus.org/articles/14/3687/2020/},
DOI = {10.5194/tc-14-3687-2020}
}
The DeepBedMap_DEM v1.1.0 dataset is available from Zenodo at https://doi.org/10.5281/zenodo.4054246. Neural network model training experiment runs are also recorded at https://www.comet.ml/weiji14/deepbedmap. Python code for the DeepBedMap model here on Github is also mirrored on Zenodo at https://doi.org/10.5281/zenodo.3752613.