dolphin-classification

This repository contains code for classifying dolphin species using deep learning techniques and supports the related publication on improving acoustic classification of two closely related species, the Atlantic bottlenose and common dolphin.

The main files are:

• code/dataprocess.R: Prepares the datasets for model training and evaluation.
• report_species.qmd: Generates the report for the species classification task. Set echo: true in YAML header and recompile to see the underlying code along the outputs in the resulting file report_species.pdf. In the section "General parameters", set rerun_CV <- TRUE to rerun the cross-validation and generate new results (the cross-validation codes are in the folder code/); it is set to FALSE by default to use the precomputed results. Similarly, set rerun_pdp <- TRUE to update partial dependence plots, which are set to FALSE by default.
• code/map_dolphins.R: Loads the metadata file from dataraw/ and generates a map of the recording locations: