AngusG / deep-learning-dreissenid

Source code for reproducing "Predicting Dreissenid Mussel Abundance using Deep Learning" by Galloway et al..

Code for: Predicting Dreissenid Mussel Abundance in Nearshore Waters using Underwater Imagery and Deep Learning

A. Galloway, D. Brunet, R. Valipour, M. McCusker, J. Biberhofer, M. K. Sobol, M. Moussa, and G. W. Taylor. (2021). Predicting Dreissenid Mussel Abundance in Nearshore Waters using Underwater Imagery and Deep Learning. Limnology and Oceanography: Methods (pending minor revisions).

Datasets can be downloaded here.

Note: Improved documentation and code cleaning for this repository is in progress.

Overview

• The predict folder contains scripts for training and evaluating deep neural networks on DS1, DS2, & DS3.
• The quadrat-extraction folder contains code for extracting the contents of quadrat frames from images and video.
• The label-me folder contains code related to dataset preparation, preprocessing, and obtaining segmentation labels.