demo.m

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Code for [1] by Sai Srivatsa R
% Email : saisrivatsan12@gmail.com
% Date : 12/09/2015
%
% Code for [2,3,4,5] by Wangjiang Zhu
% Email: wangjiang88119@gmail.com
% Date: 3/24/2014
%
% If you use this code, please cite both [1] and [2].
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% This demo shows how to use Saliency Objectness[1], as well as Saliency 
% Optimization[2], Saliency Filter[3], Geodesic Saliency[4], 
% and Manifold Ranking[5].

% [1] Sai Srivatsa R, R Venkatesh Babu. Salient Object Detection via
% Objectnes Measuew. In ICIP, 2015.

% [2] Wangjiang Zhu, Shuang Liang, Yichen Wei, and Jian Sun. Saliency
% Optimization from Robust Background Detection. In CVPR, 2014.

% [3] F. Perazzi, P. Krahenbuhl, Y. Pritch, and A. Hornung. Saliency
% filters: Contrast based filtering for salient region detection.
% In CVPR, 2012.

% [4] Y.Wei, F.Wen,W. Zhu, and J. Sun. Geodesic saliency using
% background priors. In ECCV, 2012.

% [5] C. Yang, L. Zhang, H. Lu, X. Ruan, and M.-H. Yang. Saliency
% detection via graph-based manifold ranking. In CVPR, 2013.

%%
clear, clc, 
close all
addpath('Funcs');

%% 1. Parameter Settings
doFrameRemoving = false;
useSP = true;           %You can set useSP = false to use regular grid for speed consideration
doMAEEval = true;       %Evaluate MAE measure after saliency map calculation
doPRCEval = true;       %Evaluate PR Curves after saliency map calculation

SRC = 'Data/SRC';       %Path of input images
SP = 'Data/SP';         %Path for saving superpixel index image and mean color image
RES = 'Data/Res';       %Path for saving saliency maps
srcSuffix = '.jpg';     %suffix for your input image

if ~exist(SP, 'dir')
    mkdir(SP);
end
if ~exist(RES, 'dir')
    mkdir(RES);
end
%% 2. Saliency Map Calculation
files = dir(fullfile(SRC, strcat('*', srcSuffix)));
% if matlabpool('size') <= 0
%     matlabpool('open', 'local', 8);
% end
for k=1:length(files)
    disp(k);
    srcName = files(k).name;
    noSuffixName = srcName(1:end-length(srcSuffix));
    %% Pre-Processing: Remove Image Frames
    srcImg = imread(fullfile(SRC, srcName));
    if doFrameRemoving
        [noFrameImg, frameRecord] = removeframe(srcImg, 'sobel');
        [h, w, chn] = size(noFrameImg);
    else
        noFrameImg = srcImg;
        [h, w, chn] = size(noFrameImg);
        frameRecord = [h, w, 1, h, 1, w];
    end
    
    %% Segment input rgb image into patches (SP/Grid)
    pixNumInSP = 600;                           %pixels in each superpixel
    spnumber = round( h * w / pixNumInSP );     %super-pixel number for current image
    
    if useSP
        [idxImg, adjcMatrix, pixelList] = SLIC_Split(noFrameImg, spnumber);
    else
        [idxImg, adjcMatrix, pixelList] = Grid_Split(noFrameImg, spnumber);        
    end
    %% Get super-pixel properties
    spNum = size(adjcMatrix, 1);
    meanRgbCol = GetMeanColor(noFrameImg, pixelList);
    meanLabCol = colorspace('Lab<-', double(meanRgbCol)/255);
    meanPos = GetNormedMeanPos(pixelList, h, w);
    bdIds = GetBndPatchIds(idxImg);
    colDistM = GetDistanceMatrix(meanLabCol);
    posDistM = GetDistanceMatrix(meanPos);
    [clipVal, geoSigma, neiSigma] = EstimateDynamicParas(adjcMatrix, colDistM);
    
    %% Saliency Objectness
    [bgProb, bdCon, bgWeight] = EstimateBgProb(colDistM, adjcMatrix, bdIds, clipVal, geoSigma);
    wCtr = SaliencyObjectness(srcName(1:end-4),h,w,idxImg,pixelList,adjcMatrix,colDistM,clipVal);
    optwCtr = SaliencyOptimization(adjcMatrix, bdIds, colDistM, neiSigma, bgWeight, wCtr);
    
    %Uncomment the next lines to store foreground weights maps
    %smapName=fullfile(RES, strcat(noSuffixName, '_SInit.png'));
    %SaveSaliencyMap(wCtr, pixelList, frameRecord, smapName, true);
 
    smapName=fullfile(RES, strcat(noSuffixName, '_SObj.png'));
    SaveSaliencyMap(optwCtr, pixelList, frameRecord, smapName, true);  
   
    %% Saliency Optimization 
    wCtr = CalWeightedContrast(colDistM, posDistM, bgProb);
    optwCtr = SaliencyOptimization(adjcMatrix, bdIds, colDistM, neiSigma, bgWeight, wCtr);
    
    smapName=fullfile(RES, strcat(noSuffixName, '_wCtr_Optimized.png'));
    SaveSaliencyMap(optwCtr, pixelList, frameRecord, smapName, true);
    
    %% Saliency Filter
    [cmbVal, contrast, distribution] = SaliencyFilter(colDistM, posDistM, meanPos);
    
    smapName=fullfile(RES, strcat(noSuffixName, '_SF.png'));
    SaveSaliencyMap(cmbVal, pixelList, frameRecord, smapName, true);    

    %% Geodesic Saliency
    geoDist = GeodesicSaliency(adjcMatrix, bdIds, colDistM, posDistM, clipVal);
    
    smapName=fullfile(RES, strcat(noSuffixName, '_GS.png'));
    SaveSaliencyMap(geoDist, pixelList, frameRecord, smapName, true);
    
    %% Manifold Ranking
    [stage2, stage1, bsalt, bsalb, bsall, bsalr] = ManifoldRanking(adjcMatrix, idxImg, bdIds, colDistM);
    
    smapName=fullfile(RES, strcat(noSuffixName, '_MR_stage2.png'));
    SaveSaliencyMap(stage2, pixelList, frameRecord, smapName, true);

end

%% 3. Evaluate MAE
if doMAEEval
    GT = 'Data/GT';
    gtSuffix = '.jpg';
    CalMeanMAE(RES, '_wCtr_Optimized.png', GT, gtSuffix);
    CalMeanMAE(RES, '_SF.png', GT, gtSuffix);
    CalMeanMAE(RES, '_GS.png', GT, gtSuffix);
    CalMeanMAE(RES, '_MR_stage2.png', GT, gtSuffix);
    CalMeanMAE(RES, '_SObj.png', GT, gtSuffix);
end

%% 4. Evaluate PR Curve
if doPRCEval
    GT = 'Data/GT';
    gtSuffix = '.jpg';
    figure, hold on;
    DrawPRCurve(RES, '_wCtr_Optimized.png', GT, gtSuffix, true, true, 'r');
    DrawPRCurve(RES, '_SF.png', GT, gtSuffix, true, true, 'g');
    DrawPRCurve(RES, '_GS.png', GT, gtSuffix, true, true, 'b');
    DrawPRCurve(RES, '_MR_stage2.png', GT, gtSuffix, true, true, 'k');
    DrawPRCurve(RES, '_SObj.png', GT, gtSuffix, true, true, 'cy');
    hold off;
    grid on;
    lg = legend({'wCtr\_opt'; 'SF'; 'GS'; 'MR';'Ours'});
    set(lg, 'location', 'southwest');
end