Baobab Detection and Analysis to Evaluate Anthropogenic Legacies in Eastern Africa

This data is related to a paper which evaluated different methods for detecting baobab trees in drone imagery, with the aim of comparing them to archaeological sites in eastern Africa.

This dataset contains:

1. An ArcGIS Pro geodatabase, with a shapefile of hand-digitized baobab locations for Unguja Island, shapefile data of archaeological sites, shapefiles of settlement data digitized from a 1907 map of the island, squares showing seasonal data for the ZMI imagery, and an outline of the island created from SRTM elevation data, shapefiles and rasters of the automated detection training and testing files, and the results of automated detection using the Max Likelihood classifier, sorted by circularity and pixel size.
2. Excel files showing the nearness analysis for baobab trees, with the raw data and tables available
3. Excel files with confusion matrices for the automated detection in different seasons
4. Python code used for the tree detection viewer
5. A ReadMe file with all workflows described in detail