sediment-ml

Nearshore image classification

Getting Started

• Install poetry, a python package and virtual environment manager
  • On macOS, in the terminal, run the command brew install poetry (assuming homebrew is installed. If not, install homebrew)
  • General installation instructions: install poetry
• Clone this repository by running git clone git@github.com:madelinemberger/sediment-ml.git or downloading the zip
• Navigate into this repository
• Run the command poetry install to set up the virtual environment for this repository

Running scripts

• After setting up the environment with poetry (see Getting Started), you can run a script like this:
  • run the command poetry run python path/to/script.py

Running notebooks

• After setting up the environment, you can launch Jupyter Notebook from within the environment like this:
  • run the command poetry run jupyter notebook

Script Information

scripts/coastline_tile_pipeline.py

This script does the following: Given a .tif file and a shapefile bundle with coastline information that spans the region of the .tif file, this script will

• tile the .tif into NxN regions (default=512x512 pixels);
• determine which of these regions intersect the coastline;
• convert the intersecting regions to PNGs and save these files;
• create a map image showing the intersecting tiles as PNGs and the coastline contour, along with the .tif boundary

Example output from this script can be seen in examples/coastline_tiles_plot.png and below:

scripts/coastline_tile_pipeline_adjacent.py

This script does the following: Given a .tif and a shapefile bundle with coastline information that spans the region of the .tif file, this script will

• tile the .tif in NxN regions (default=512x512 pixels),
• determine which of these regions intersect the coastline,
• determine which of the regions adjacent to intersecting regions are not contained within the landmass,
• convert the intersecting regions and adjacent, non-land regions to PNGs and save these files,
• create a map image showing the intersecting tiles as red-boundary PNGs, the adjacent non-land tiles as orange-boundary PNGs, and the coastline contour as a blue line, along with the .tif boundary as a black line

Example output from this script can be seen in examples/coastline_tiles_plot_extended.png and below: