Kaggle Understanding Cloud Organization

Kaggle competition to classify cloud structures from satellite images.

Project page on GitHub.
Poster version presented at SMILES 2020 and DS³ 2021 summer schools.
Slides version presented at DS³ 2021 winter school.
Slides version presented at EEML 2020 summer school. And Teaser.

Co-authors and teammates:

Citation

Please use this bibtex if you want to cite this work in your publications:

@misc{kirill_a_vishniakov_2021_5055450,
  author       = {Kirill A. Vishniakov and Mukharbek Organokov},
  title        = {{Segmentation of cloud patterns from satellite images to improve climate models}},
  month        = jul,
  year         = 2021,
  note         = {{Code is available at GitHub: 
                   https://github.com/LightnessOfBeing/kaggle-understanding-cloud-organization}},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.5055450},
  url          = {https://doi.org/10.5281/zenodo.5055450}
}

Abstract

Climate change has been at the top of our minds and at the forefront of important political decision-making for many years. Classification of different types of clouds is substantial for understanding climate change. Human ability to identify patterns is limited and murky boundaries between different forms of clouds lead to obstacles in traditional rule-based algorithms cloud features separation. In these situations, machine learning techniques, particularly deep learning, have demonstrated their ability to mimic the human capacity for identifying patterns in the clouds using satellite images. This work focuses on the segmentation of four subjective patterns of clouds organization: Sugar, Flower, Fish, Gravel.