Skip to content

Latest commit

 

History

History
36 lines (26 loc) · 1.77 KB

README.md

File metadata and controls

36 lines (26 loc) · 1.77 KB

MapInWild

This repository contains the code for the paper "MapInWild: A Remote Sensing Dataset to Answer the Question What Makes Nature Wild".

MapInWild dataset is available here. The Python API of the Harvard Dataverse can be used for bulk actions.

See the folders segmentation and sensitivity for the SEMANTIC SEGMENTATION and the SCENE CLASSIFICATION AND SENSITIVITY ANALYSIS experiments, respectively.

A sample from MapInWild dataset is shown below. The first row: four-season Sentinel-2 patches, second row: Sentinel-1 image, ESA WorldCover map, VIIRS Nighttime Day/Night band, and World Database of Protected Areas (WDPA) annotation.

alt text

The files in the dataset are named after the ID of the protected area they contain.

Batch visalualizations from MapInWild dataset:

  • 256 x 256 patches alt text
  • 1920 x 1920 full-size patches alt text

Citation

@ARTICLE{10089830,
  author={Ekim, Burak and Stomberg, Timo T. and Roscher, Ribana and Schmitt, Michael},
  journal={IEEE Geoscience and Remote Sensing Magazine}, 
  title={MapInWild: A remote sensing dataset to address the question of what makes nature wild [Software and Data Sets]}, 
  year={2023},
  volume={11},
  number={1},
  pages={103-114},
  doi={10.1109/MGRS.2022.3226525}}