Datasets for paper on Carolina Bay Detection
Link to directory here.
In the 'Spreadsheets' directory, you can find csvs containing the following:
Precision and recall data used to construct the PR curves.
Data used for validation on the Delaware datasets (in Spreadsheets/Validation_Delaware)
Filtered Carolina Bay detections joined with morphometry, land-use, and surface hydrologic characteristics (in Spreadsheets/Full_ACP).
Sedimentological data is available at github.com/mlundine/CarolinaBaySedimentology.
Data for the tile size and overlap experiment (in Spreadsheets/Tile_Size_Overlap_Experiment).
Data for the principal component analysis on topo metrics.
Link to directory here.
In the 'GIS_Files' directory, you can find shapefiles of Carolina Bay detections obtained through running the trained Faster R-CNN
model on the Atlantic Coastal Plain DEM tiled at various footprints and then running them through the PAEK smoothing algorithm (in GIS_Files/Final_Detections).
You can also find the unfiltered detections from various tile sizes as shapefiles (in GIS_Files/Multi_Scale_Unfiltered).
You can also find the shapefiles constructed from the tile/overlap experiment (in GIS_Files/Tile_Overlap_Experiment).
Link to directory here.
In the 'Trained_Models' directory, you can find the trained Faster R-CNN (in Trained_Models/frcnn_inference_graph), Mask R-CNN (in Trained_Models/mrcnn_inference_graph), and yolov5 (in Trained_Models/best_carolinaBays) models. This repo has a GUI that you can use to run the models: github.com/mlundine/tensorflow_app
Instructions on how to make the mosaic of the Atlantic Coastal Plain DEM as well as the data needed are available here.