Skip to content

Latest commit

 

History

History
29 lines (18 loc) · 1.31 KB

README.md

File metadata and controls

29 lines (18 loc) · 1.31 KB
Raw

interpreting-unet

Interpreting a U-Net used for coastal water body segmentation using permutation importance

This repository contains the code required to reproduce the results in the conference paper:

To update

This code is only for academic and research purposes. Please cite the above paper if you intend to use whole/part of the code.

Data Files

We have used the following dataset in our analysis:

  1. Sentinel-2 Water Edges Dataset (SWED) from UK Hydrographic Office.

The data is available under the Geospatial Commission Data Exploration license.

Code Files

You can find the following files in the src folder:

  • train_unet.py Code used to train the semantic segmentation model
  • analysis_of_indices.ipynb The main analysis file used to calculate permutation importance and create all figures in the research paper.
  • utils.py Helper file containing functions used to perform the analysis in the main analysis file.

Result Files

You can find the following files used in the analysis:

  • UNET-SCALE-13MAR23.pth Final U-Net model which is interpreted in this analysis
  • Metrics_28MAR23.csv Contains the permutation importance for different spectral bands and combinations of bands