Interpreting a Semantic Segmentation Model for Coastline Detection

TL;DR

This paper interprets a deep learning model for coastline detection, revealing spectral band importance and suggesting potential simplifications to improve model efficiency.

Contribution

It introduces a permutation importance method to identify key spectral bands in coastline segmentation, highlighting the significance of NIR and Water Vapour bands.

Findings

NIR band is most important, decreasing accuracy by 38.12% when permuted.

Water Vapour band significantly impacts model accuracy, suggesting its usefulness.

Several bands can be excluded without affecting performance, reducing model complexity.

Abstract

We interpret a deep-learning semantic segmentation model used to classify coastline satellite images into land and water. This is to build trust in the model and gain new insight into the process of coastal water body extraction. Specifically, we seek to understand which spectral bands are important for predicting segmentation masks. This is done using a permutation importance approach. Results show that the NIR is the most important spectral band. Permuting this band lead to a decrease in accuracy of 38.12 percentage points. This is followed by Water Vapour, SWIR 1, and Blue bands with 2.58, 0.78 and 0.19 respectively. Water Vapour is not typically used in water indices and these results suggest it may be useful for water body extraction. Permuting, the Coastal Aerosol, Green, Red, RE1, RE2, RE3, RE4, and SWIR 2 bands did not decrease accuracy. This suggests they could be excluded from future model builds reducing complexity and computational requirements.