function [Xtrain , Xtest , stat] = generate_data_cascade(options , Xfa )
%
% Generate i.i.d positive features from positives examples stacked in a 3D
% tensor matrix Xpos and negative samples from negatives picts stores in
% negatives folder
%
% Usage
% ------
%
% [Xtrain , Xtest , stat] = generate_data_cascade(options , Xfa )
%
% Inputs
% -------
%
% options Options struture (see train_cascade function)
% Xfa Previous Falses alarms detection matrix (Ny x Nx x Nfa)
%
% Outputs
% -------
%
% Xtrain Generated train data (Ny x Nx x (options.Npostrain+options.Nnegtrain))
% Xtest Generated test data (Ny x Nx x (options.Npostest+options.Nnegtest))
% stat Number of positives and negatives generated for both train and test set
%
%
%
% Author : Sbastien PARIS : sebastien.paris@lsis.org
% ------- Date : 03/1/2011
%
%
if(options.resetseed)
RandStream.setDefaultStream(RandStream.create('mt19937ar','seed',options.seed));
end
Ntotal = options.Ntotal;
X = zeros(options.ny , options.nx , Ntotal , 'uint8');
stat = zeros(5,1);
%% Generate Npos Positives examples from Xpos without replacement %%
Npostotal = options.Npos;
indexpos = randperm(Npostotal);
nb_stage = length(options.m);
directory = [];
for i = 1:length(options.negext)
directory = [directory ; dir(fullfile(options.positives_path , ['*.' options.posext{i}]))] ;
end
nbpos_photos = length(directory);
index_posphotos = randperm(nbpos_photos);
copos = 1;
co = 1;
cophotos = 1;
h = waitbar(0,sprintf('Generating positives, n^o stage = %d' , nb_stage));
set(h , 'name' , sprintf('Generating positives, n^o stage = %d' , nb_stage))
if(options.preview)
fig = figure(1);
set(fig,'doublebuffer','on');
colormap(gray)
end
while((copos <= options.Npos) && (co <= Npostotal) )
if(~isempty(options.param))
if(cophotos < nbpos_photos)
cophotos = cophotos + 1;
else
cophotos = 1;
index_posphotos = randperm(nbpos_photos);
end
Ipos = imread(fullfile(options.positives_path , directory(index_posphotos(cophotos)).name));
[Ny , Nx , Nz] = size(Ipos);
if(Nz == 3)
Ipos = rgb2gray(Ipos);
end
%% Resize to training %%
Ipos = imresize(Ipos , options.dimsItraining);
%% Standardize image %%
if(options.standardize)
dIpos = double(Ipos);
stdIpos = std(dIpos(:));
if(stdIpos~=0)
dIpos = dIpos/stdIpos;
end
minIpos = min(dIpos(:));
maxIpos = max(dIpos(:));
tempX = uint8(floor(255*(dIpos-minIpos)/(maxIpos-minIpos)));
end
%% Rotation %%
if(rand < options.probarotIpos)
tempX = fast_rotate(tempX , options.m_angle + options.std_angle*randn(1,1));
end
%% preview image %%
if(options.typefeat == 0)
[fxtempX , ytempX] = eval_haar(tempX , options);
if(ytempX == 1)
X(: , : , copos) = tempX;
copos = copos + 1;
if(options.preview)
imagesc(tempX);
drawnow
end
end
else
[fxtempX , ytempX] = eval_mblbp(tempX , options);
if(ytempX == 1)
X(: , : , copos) = tempX;
copos = copos + 1;
if(options.preview)
imagesc(tempX);
drawnow
end
end
end
else
if(cophotos < nbpos_photos)
cophotos = cophotos + 1;
else
cophotos = 1;
index_posphotos = randperm(nbpos_photos);
end
Ipos = imread(fullfile(options.positives_path , directory(index_posphotos(cophotos)).name));
[Ny , Nx , Nz] = size(Ipos);
if(Nz == 3)
Ipos = rgb2gray(Ipos);
end
%% Resize to training %%
Ipos = imresize(Ipos , options.dimsItraining);
%% Standardize image %%
if(options.standardize)
dIpos = double(Ipos);
stdIpos = std(dIpos(:));
if(stdIpos~=0)
dIpos = dIpos/stdIpos;
end
minIpos = min(dIpos(:));
maxIpos = max(dIpos(:));
tempX = uint8(floor(255*(dIpos-minIpos)/(maxIpos-minIpos)));
end
%% Rotation image %%
if(rand < options.probarotIpos)
tempX = fast_rotate(tempX , options.m_angle + options.std_angle*randn(1,1));
end
%% preview image %%
if(options.preview)
imagesc(tempX);
drawnow
end
X(: , : , copos) = tempX;
copos = copos + 1;
end
co = co + 1;
waitbar(copos/options.Npos , h , sprintf('#pos/#generated = %d/%d, P_d = %5.4f' , copos - 1 , co - 1 , (copos - 1)/(co - 1)));
end
close(h)
stat(1) = copos - 1;
stat(2) = co - 1;
%% Generate Nneg Negatives examples from Negatives images with replacement %%
if(options.usefa == 1)
prenegnum = size(Xfa,3);
X(:,:,options.Npos+1:prenegnum+options.Npos) = Xfa;
coneg = prenegnum+options.Npos+1;
else
prenegnum = 0;
coneg = options.Npos+1;
end
cophotos = 1;
directory = [];
for i = 1:length(options.negext)
directory = [directory ; dir(fullfile(options.negatives_path , ['*.' options.negext{i}]))] ;
end
nbneg_photos = length(directory);
index_negphotos = randperm(nbneg_photos);
Ineg = imread(fullfile(options.negatives_path , directory(index_negphotos(cophotos)).name));
[Ny , Nx , Nz] = size(Ineg);
if(Nz == 3)
Ineg = rgb2gray(Ineg);
end
maxNyNx = max([Ny Nx]);
if(maxNyNx > options.negmax_size)
ratio = options.negmax_size/maxNyNx;
Ny = ceil(ratio*Ny);
Nx = ceil(ratio*Nx);
Ineg = imresize(Ineg , [Ny Nx]);
end
if(min(Ny,Nx) < options.scalemax*max(options.ny,options.nx))
scalemax = min(Ny,Nx)/max(options.ny,options.nx);
else
scalemax = options.scalemax;
end
if(options.preview)
imagesc(Ineg);
end
h = waitbar(0,sprintf('Generating negatives, n^o stage = %d' , nb_stage));
set(h , 'name' , sprintf('Generating negatives, n^o stage = %d' , nb_stage))
co = 1;
while(coneg <= Ntotal)
if(~isempty(options.param))
if(rand < options.probaswitchIneg)
if(cophotos < nbneg_photos)
cophotos = cophotos + 1;
else
cophotos = 1;
index_negphotos = randperm(nbneg_photos);
end
Ineg = imread(fullfile(options.negatives_path , directory(index_negphotos(cophotos)).name));
[Ny , Nx , Nz] = size(Ineg);
if(Nz == 3)
Ineg = rgb2gray(Ineg);
end
if(min(Ny,Nx) < options.scalemax*max(options.ny,options.nx))
scalemax = min(Ny,Nx)/max(options.ny,options.nx);
end
end
%% Scaling Negatives %%
scale = options.scalemin + (scalemax-options.scalemin)*rand;
nys = round(options.ny*scale);
nxs = round(options.nx*scale);
y = ceil(1 + (Ny - nys - 1)*rand);
x = ceil(1 + (Nx - nxs - 1)*rand);
tempX = imresize(Ineg(y:y+nys-1 , x:x+nxs-1) , [options.ny , options.nx]);
%% standardize Negatives %%
if(options.standardize)
dIneg = double(tempX);
stdIneg = std(dIneg(:));
if(stdIneg~=0)
dIneg = dIneg/stdIneg;
end
minIneg = min(dIneg(:));
maxIneg = max(dIneg(:));
tempX = uint8(floor(255*(dIneg-minIneg)/(maxIneg-minIneg)));
end
if(options.typefeat == 0)
[fxtempX , ytempX] = eval_haar(tempX , options);
if(std(double(tempX(:)),1)~= 0)
if(ytempX == 1)
X(: , : , coneg ) = tempX;
coneg = coneg + 1;
if(options.preview)
imagesc(tempX);
drawnow
end
end
co = co + 1;
end
else
[fxtempX , ytempX] = eval_mblbp(tempX , options);
if(std(double(tempX(:)),1)~= 0)
if(ytempX == 1)
X(: , : , coneg ) = tempX;
coneg = coneg + 1;
if(options.preview)
imagesc(tempX);
drawnow
end
end
co = co + 1;
end
end
else
if(rand < options.probaswitchIneg)
if(cophotos < nbneg_photos)
cophotos = cophotos + 1;
else
cophotos = 1;
index_negphotos = randperm(nbneg_photos);
end
Ineg = imread(fullfile(options.negatives_path , directory(index_negphotos(cophotos)).name));
[Ny , Nx , Nz] = size(Ineg);
if(Nz == 3)
Ineg = rgb2gray(Ineg);
end
if(min(Ny,Nx) < options.scalemax*max(options.ny,options.nx))
scalemax = min(Ny,Nx)/max(options.ny,options.nx);
end
end
scale = options.scalemin + (scalemax-options.scalemin)*rand;
nys = round(options.ny*scale);
nxs = round(options.nx*scale);
y = ceil(1 + (Ny - nys - 1)*rand);
x = ceil(1 + (Nx - nxs - 1)*rand);
Itemp = Ineg(y:y+nys-1 , x:x+nxs-1);
tempX = imresize(Itemp , [options.ny , options.nx]);
if(options.standardize)
dIneg = double(tempX);
stdIneg = std(dIneg(:));
if(stdIneg~=0)
dIneg = dIneg/stdIneg;
end
minIneg = min(dIneg(:));
maxIneg = max(dIneg(:));
tempX = uint8(floor(255*(dIneg-minIneg)/(maxIneg-minIneg)));
end
if(std(double(tempX(:)),1)~= 0)
X(: , : , coneg ) = tempX;
coneg = coneg + 1;
co = co + 1;
if(options.preview)
imagesc(tempX);
drawnow
end
end
end
waitbar((coneg - (options.Npos+1))/options.Nneg , h , sprintf('#neg/#generated = %d/%d, P_{fa} = %5.4f, scale = %4.2f' , coneg - (options.Npos+1), co + prenegnum - 1 , (coneg - (options.Npos+1))/(co + prenegnum - 1) , scale));
end
close(h)
stat(3) = coneg-1-stat(1);
stat(4) = co-1;
if(options.usefa == 1)
stat(5) = prenegnum;
else
stat(5) = 0;
end
Xtrain = X(:,:,[1:options.Npostrain,options.Npos+1:options.Npos+options.Nnegtrain]);
Xtest = X(:,:,[options.Npostrain+1:options.Npos,options.Npos+options.Nnegtrain+1:Ntotal]);
