sceneTrain.m

%% Train KNN model for pitching scene classification
pitchingFiles = dir('pitching/*.jpg'); 
pitchingNumFiles = length(pitchingFiles);
pitchingFeatures = zeros(pitchingNumFiles, 12);
pitchingResponses = ones(pitchingNumFiles, 1);
for k = 1:pitchingNumFiles
    im = imread(strcat('pitching/', pitchingFiles(k).name));
    feature = sceneFeature(im);
    pitchingFeatures(k, :) = feature;
end

nonPitchingFiles = dir('nonpitching/*.jpg'); 
nonPitchingNumFiles = length(nonPitchingFiles);
nonPitchingFeatures = zeros(nonPitchingNumFiles, 12);
nonPitchingResponses = zeros(nonPitchingNumFiles, 1);
for k = 1:nonPitchingNumFiles
    im = imread(strcat('nonpitching/', nonPitchingFiles(k).name));
    feature = sceneFeature(im);
    nonPitchingFeatures(k, :) = feature;
end

features = vertcat(pitchingFeatures, nonPitchingFeatures);
responses = vertcat(pitchingResponses, nonPitchingResponses);
sceneModel = fitcknn(features,responses,'NumNeighbors',3,'DistanceWeight','squaredinverse');
% save trained model
save('sceneModel.mat', 'sceneModel');