Skip to content

Commit

Permalink
correct flow_mean.mat
Browse files Browse the repository at this point in the history
  • Loading branch information
@wanglimin
wanglimin committed Jul 29, 2015
1 parent 85f71af commit 549e93b
Show file tree
Hide file tree
Showing 2 changed files with 88 additions and 86 deletions.
174 changes: 88 additions & 86 deletions extract_fv.m
View file
Original file line number Diff line number Diff line change
Expand Up @@ -25,98 +25,100 @@ function extract_fv(index, power, layer, tag ,dim1, norm)
foldername = setdiff(foldername,{'.','..'});

for i = index
% if ~exist([path6,foldername{i}],'dir')
% mkdir([path6,foldername{i}]);
% end
if ~exist([path7,foldername{i}],'dir')
% if ~exist([path6,foldername{i}],'dir')
% mkdir([path6,foldername{i}]);
% end
if ~exist([path7,foldername{i}],'dir')
mkdir([path7,foldername{i}]);
end
if ~exist([path8,foldername{i}],'dir')
end
if ~exist([path8,foldername{i}],'dir')
mkdir([path8,foldername{i}]);
end
if ~exist([path9,foldername{i}],'dir')
end
if ~exist([path9,foldername{i}],'dir')
mkdir([path9,foldername{i}]);
end
if ~exist([path10,foldername{i}],'dir')
end
if ~exist([path10,foldername{i}],'dir')
mkdir([path10,foldername{i}]);
end
end
filelist = dir([path3,foldername{i},'/*.mat']);
tic;
tic;
for j = 1:length(filelist)

% feature = load([path1,foldername{i},'/',filelist(j).name]);
% feature = double(feature.idt_cnn_feature.^power);
%
% if ~isempty(feature)
% feature = bsxfun(@minus,feature,PCA.mu);
% feature = PCA.U(:,1:dim1)'*feature;
% feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
% coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
% else
%
% coding = zeros(1,2*num*dim1);
% end
% save([path6,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path2,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path7,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path3,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path8,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path4,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path9,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path5,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path10,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

end
toc;
if ~exist([path10,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'file')

% feature = load([path1,foldername{i},'/',filelist(j).name]);
% feature = double(feature.idt_cnn_feature.^power);
%
% if ~isempty(feature)
% feature = bsxfun(@minus,feature,PCA.mu);
% feature = PCA.U(:,1:dim1)'*feature;
% feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
% coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
% else
%
% coding = zeros(1,2*num*dim1);
% end
% save([path6,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path2,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path7,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path3,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path8,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path4,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path9,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');

feature = load([path5,foldername{i},'/',filelist(j).name]);
feature = double(feature.idt_cnn_feature.^power);

if ~isempty(feature)
feature = bsxfun(@minus,feature,PCA.mu);
feature = PCA.U(:,1:dim1)'*feature;
feature = bsxfun(@rdivide,feature,sqrt(PCA.vars(1:dim1)));
coding = vl_fisher(feature,GMM.means,GMM.covariances,GMM.priors);
else

coding = zeros(1,2*num*dim1);
end
save([path10,foldername{i},'/',filelist(j).name(1:end-4),'_pca_',num2str(dim1),'_power_',num2str(power),'.mat'],'coding');
end

end
toc;
end

end
Expand Down
Binary file modified flow_mean.mat
View file
Binary file not shown.

0 comments on commit 549e93b

Please sign in to comment.