The training data used for the analysis in the manuscript "Automatic Classification of Biological Targets in a Tidal Channel using a Multibeam Sonar," by Emma Cotter and Brian Polagye, are included in the files '2017.mat,' '2019a.mat,' and '2019b.mat.' These files correspond to the 2017, 2019a, and 2019b trarining data sets discussed in the manuscript, respectively. A fourth file, 'features.mat', contains the list of features included for each target track in the training data and their corresponding indices in the training data. Features are described in deatail in Appendix A of the manuscript. 

Each training data file includes the matrix X, containing all calculated features for each target track in the training data set, and the vector y, that contains the reviewer-assigned class for each target track (using the abbreviations in Table 1 of the manuscript). For the 2019a and 2019b data sets, two sets of X and y are provided. 'track_end' includes only the set of features calculated at the end of each target track, and 'all_points' contains the set of features calculated at each timestep during real-time classification. The training data are not normalized (equation 1 in the manuscript has not yet been applied). 

'features.mat' contains a structure, 'features', that lists the indices in the matrix X that correspond to each feature. For example, I, the 75th percentile intensity, is the first column of each matrix X.