script_experiments_UNIX.m

warning('off','all');

% Datasets names
datasetsNames = {...
     'DRIVE' ...
     'STARE-A' ...
     'STARE-B' ...
     'CHASEDB1-B' ...
     'CHASEDB1-A' ...
     'HRF' ...
    };

% Flag indicating if the value of C is going to be tuned according to the
% validation set
learnC = 1;
% CRF versions that are going to be evaluated
crfVersions = {'up','fully-connected'};
% C values
cValue = 10^2;

% Root dir where the data sets are located
rootDatasets = '/home/ignacioorlando/_tmi_experiments/';

% Root folder where the results are going to be stored
rootResults = '/home/ignacioorlando/results-last-all';





% Creating data set paths
datasetsPaths = cell(length(datasetsNames), 1);
for i = 1 : length(datasetsNames)
    datasetsPaths{i} = strcat(rootDatasets, datasetsNames{i});
end

% Creating results paths
resultsPaths = cell(length(datasetsNames), 1);
for i = 1 : length(datasetsNames)
    resultsPaths{i} = strcat(rootResults, filesep, datasetsNames{i});
end





for experiment = 1 : length(datasetsNames)

    for crfver = 1 : length(crfVersions)
               
        % Get the configuration
        [config] = getConfiguration_GenericDataset_UNIX(datasetsNames{experiment}, datasetsPaths{experiment}, resultsPaths{experiment}, learnC, crfVersions{crfver}, cValue);

        root = config.resultsPath;

        % Code name of the expected files
        pairwisedeviations = strcat(config.training_data_path, filesep, 'pairwisedeviations.mat');

        % Open images, labels and masks on the training set
        [trainingdata.images, trainingdata.labels, trainingdata.masks, trainingdata.numberOfPixels] = openLabeledData(config.training_data_path, config.preprocessing);
        % Open images, labels and masks for the validation set
        [validationdata.images, validationdata.labels, validationdata.masks, validationdata.numberOfPixels] = openLabeledData(config.validation_data_path, config.preprocessing);

        % If the pairwise deviation file does not exist
        if (exist(pairwisedeviations, 'file')~=2)
            % Compute all possible features
            [allfeatures, numberOfDeviations] = extractFeaturesFromImages(trainingdata.images, trainingdata.masks, config, ones(size(config.features.numberFeatures)), false);
            % Compute pairwise deviations
            pairwiseDeviations = getPairwiseDeviations(allfeatures, numberOfDeviations);
            % Save pairwise deviations
            save(pairwisedeviations, 'pairwiseDeviations');
        else
            % Load pairwise deviations
            load(pairwisedeviations); 
        end

        % Assign precomputed deviations to the param struct
        config.features.pairwise.pairwiseDeviations = pairwiseDeviations;
        clear 'pairwiseDeviations';

        % Extract unary features
        fprintf('Computing unary features\n');
        % Compute unary features on training data
        [trainingdata.unaryFeatures, config.features.unary.unaryDimensionality] = ...
            extractFeaturesFromImages(trainingdata.images, ...
                                      trainingdata.masks, ...
                                      config, ...
                                      config.features.unary.unaryFeatures, ...
                                      true);
        % Compute unary features on validation data
        validationdata.unaryFeatures = extractFeaturesFromImages(validationdata.images, ...
                                                                 validationdata.masks, ...
                                                                 config, ...
                                                                 config.features.unary.unaryFeatures, ...
                                                                 true);

        % Compute pairwise features on training data
        % Extract pairwise features
        fprintf('Computing pairwise features\n');
        [pairwisefeatures, config.features.pairwise.pairwiseDimensionality] = ...
            extractFeaturesFromImages(trainingdata.images, ...
                                      trainingdata.masks, ...
                                      config, ...
                                      config.features.pairwise.pairwiseFeatures, ...
                                      false);
        config.features.pairwise.pairwiseDeviations = config.features.pairwise.pairwiseDeviations(generateFeatureFilter(config.features.pairwise.pairwiseFeatures, config.features.pairwise.pairwiseFeaturesDimensions));
        trainingdata.pairwiseKernels = getPairwiseFeatures(pairwisefeatures, config.features.pairwise.pairwiseDeviations);

        % Compute pairwise features on validation data
        pairwisefeatures = extractFeaturesFromImages(validationdata.images, ...
                                                     validationdata.masks, ...
                                                     config, ...
                                                     config.features.pairwise.pairwiseFeatures, ...
                                                     false);
        validationdata.pairwiseKernels = getPairwiseFeatures(pairwisefeatures, config.features.pairwise.pairwiseDeviations);

        % Filter the value of theta_p
        config.theta_p.finalValues = ...
            config.theta_p.values(generateFeatureFilter(config.features.pairwise.pairwiseFeatures, config.features.pairwise.pairwiseFeaturesDimensions));

        % Train with this configuration and return the model
        [model, config.qualityOverValidation, config] = learnCRFPotentials(config, trainingdata, validationdata);


        % Open test data
        [testdata.images, testdata.labels, testdata.masks, testdata.numberOfPixels] = openLabeledData(config.test_data_path, config.preprocessing);

        % Extract unary features
        fprintf(strcat('Computing unary features\n'));
        [testdata.unaryFeatures, config.features.unary.unaryDimensionality] = ...
            extractFeaturesFromImages(testdata.images, ...
                                      testdata.masks, ...
                                      config, ...
                                      config.features.unary.unaryFeatures, ...
                                      true);

        % Extract pairwise features
        fprintf(strcat('Computing pairwise features\n'));
        [pairwisefeatures, config.features.pairwise.pairwiseDimensionality] = ...
            extractFeaturesFromImages(testdata.images, ...
                                      testdata.masks, ...
                                      config, ...
                                      config.features.pairwise.pairwiseFeatures, ...
                                      false);

        % Compute the pairwise kernels
        fprintf(strcat('Computing pairwise kernels\n'));
        testdata.pairwiseKernels = getPairwiseFeatures(pairwisefeatures, config.features.pairwise.pairwiseDeviations);

        % Segment test data to evaluate the model
        [results.segmentations, results.qualityMeasures] = getBunchSegmentations2(config, testdata, model);

        results.table = [results.qualityMeasures.se, results.qualityMeasures.sp, results.qualityMeasures.acc, results.qualityMeasures.precision, results.qualityMeasures.recall, results.qualityMeasures.fMeasure, results.qualityMeasures.matthews];
        disp(strcat('Se = ', num2str(mean(results.qualityMeasures.se))));
        disp(strcat('Sp = ', num2str(mean(results.qualityMeasures.sp))));
        disp(strcat('fMeasure = ', num2str(mean(results.qualityMeasures.fMeasure))));
        disp(strcat('matthews = ', num2str(mean(results.qualityMeasures.matthews))));

        SaveSegmentations(root, config, results, model);
        
    end
    
end