pyShore is an open-source toolkit for mapping shoreline structures with a deep learning framework pre-trained on high-resolution orthoimagery. The following flowchart outlines the use of this tool in ArcGIS Pro:
- Clone the pyShore GitHub repository, or save the pyShoreArcGIS and weights folder in your defined project working directory.
- In ArcGIS Pro, turn on Catalog Pane under the "View" tab (section A in chart below).
- Go to "Catalog" and add the pyShore toolbox by clicking "Add Toolbox" (section B).
- The "CoastalStructurePrediction" toolkit is added to your environment and ready to use (sections C and D).
There are four inputs required to run the toolkit:
ProjectDir: a working folder in which to save all processing and results data.
ImageDir: a user defined folder with all source imagery in a georeferenced (geoTiff) format.
GeoPath: a user defined single geometry file (i.e., shapefile) that defines the geographic locations of shoreline within the provided imagery.
BufferDist: a user defined distance to buffer shorelines, the distance from the shoreline that should be extracted from the imagery to serve as the target for classification.
- torch: 1.12.0
- cuda: 11.2
- python: 3.7.10
ArcGIS Pro 2.3.0
Your ArcGIS Pro environment should have the deep learning framework set up. To set up your environment, please follow: https://pro.arcgis.com/en/pro-app/2.6/tool-reference/image-analyst/pdf/deep_learning_install.pdf
You will also need the geopandas and rasterio libraries. Some open-source Python libraries are not automatically installed in ArcGIS Pro. To install geopandas, activate the virtual environment you created from above tutorial and run: $conda install geopandas libtiff=4.0.10 rasterio can be installed through conda or pip.
Please consider citing the our paper if you find it helpful. Thank you!
@article{LV2023,
title = {pyShore: A deep learning toolkit for shoreline structure mapping with high-resolution orthographic imagery and convolutional neural networks},
journal = {Computers & Geosciences},
volume = {171},
pages = {105296},
year = {2023},
issn = {0098-3004},
doi = {https://doi.org/10.1016/j.cageo.2022.105296},
url = {https://www.sciencedirect.com/science/article/pii/S009830042200245X},
author = {Zhonghui Lv and Karinna Nunez and Ethan Brewer and Dan Runfola}