sentinel2-vegetation-line

Detecting sub-pixel changes in the coastal vegetation line with Sentinel-2 imagery

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

[coming soon]

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. The Sentinel-2 Irish Vegetation Edge (SIVE) Dataset here.

The data is available under the Creative Commons Attribution 4.0 International license.

Code Files

• demo.ipynb demonstration of the best SimpleCNN model.

You can find the following files in the src folder:

• 0_process_rasters.ipynb Stack Sentinel-2 scenes, Guidance band and vegetation lines so they can be used to create a modelling dataset.
• 0_process_rasters_erosion.ipynb Process Sentinel-2 scenes so they can be used to calculate erosion rates.
• 1_create_model_dataset.ipynb Create training and test data for edge detection machine learning models.
• 2_model_evaluation.ipynb Produce metrics and visualisations for the performance of all edge detection models.
• 3_calculate_average_lines.ipynb Calculate the average detected vegetation line (AVDLs) and format results for DSAS.
• 4_compare_erosion_rates.ipynb Visualise and summarise the erosion rates and distance calculations from DSAS.
• 5_figures_for_paper.ipynb Additional figures for research paper
• utils.py Helper functions used to perform the analysis.
• evaluation.py Help functions used to evaluate the edge detection models.
• network_hed.py Holistic nested edge detection architecture and backbones.
• train.py Training loop and hyperparameter testing.