Table of Contents
This project contains the code for a U-Net model architecture used for sea ice charting segmentation. It also contains the names of files used for training and validation (testing). New: Loss functions added for ICLR conference paper. (31/04/2022)
This project uses Python 3.9, and requires PyTorch 1.10.0 library and its dependencies.
To install this project either download it as a .zip file and extract it into
a desired directory or clone it via the terminal or console command:
- using the HTTPS
git clone https://github.com/astokholm/AI4SeaIce.git- or SSH
git clone git@github.com:astokholm/AI4SeaIce.gitAt its current development stage, the scripts in the repository are not meant to be run as executable code, but to
provided the details about the architecture and parameters of the models used in sea ice charting experiments (see
models). In addition, loss functions have recently been added for an loss comparison paper accepted at the ICLR (see functions).
The training and validation split used in the experiments is available in datalists.
Distributed
under MIT licence. See LICENSE for more information.
E-mail: stokholm@space.dtu.dk, andrzej.kucik@esa.int
Project Link: https://github.com/astokholm/AI4SeaIce
