support for conversion of draft HH models into new format

.gitmodules

@@ -53,3 +53,6 @@
 [submodule "external/base/samplers/looplessFluxSampler"]
 	path = external/base/samplers/looplessFluxSampler
 	url = https://github.com/rmtfleming/looplessFluxSampler
+[submodule "external/base/samplers/constrained-logconcave-sampler"]
+	path = external/base/samplers/constrained-logconcave-sampler
+	url = https://github.com/Bounciness/constrained-logconcave-sampler

external/base/utilities/histogram_distance/build.m

@@ -0,0 +1,11 @@
+% Build the C/C++ files provided in the package
+%
+% @author: B. Schauerte
+% @date:   2009
+% @url:    http://cvhci.anthropomatik.kit.edu/~bschauer/
+
+cpp_files=dir('*.cpp');
+for i=1:length(cpp_files)
+  fprintf('Building %d of %d: %s\n',i,length(cpp_files),cpp_files(i).name);
+  mex(cpp_files(i).name);
+end

external/base/utilities/histogram_distance/chi_square_statistics.m

@@ -0,0 +1,55 @@
+function d=chi_square_statistics(XI,XJ)
+  % Implementation of the Chi^2 distance to use with pdist
+  % (cf. "The Earth Movers' Distance as a Metric for Image Retrieval",
+  %      Y. Rubner, C. Tomasi, L.J. Guibas, 2000)
+  %
+  % @author: B. Schauerte
+  % @date:   2009
+  % @url:    http://cvhci.anthropomatik.kit.edu/~bschauer/
+
+  % Copyright 2009 B. Schauerte. All rights reserved.
+  % 
+  % Redistribution and use in source and binary forms, with or without 
+  % modification, are permitted provided that the following conditions are 
+  % met:
+  % 
+  %    1. Redistributions of source code must retain the above copyright 
+  %       notice, this list of conditions and the following disclaimer.
+  % 
+  %    2. Redistributions in binary form must reproduce the above copyright 
+  %       notice, this list of conditions and the following disclaimer in 
+  %       the documentation and/or other materials provided with the 
+  %       distribution.
+  % 
+  % THIS SOFTWARE IS PROVIDED BY B. SCHAUERTE ''AS IS'' AND ANY EXPRESS OR 
+  % IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 
+  % WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 
+  % DISCLAIMED. IN NO EVENT SHALL B. SCHAUERTE OR CONTRIBUTORS BE LIABLE 
+  % FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
+  % CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
+  % SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR 
+  % BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 
+  % WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+  % OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+  % ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+  % 
+  % The views and conclusions contained in the software and documentation
+  % are those of the authors and should not be interpreted as representing 
+  % official policies, either expressed or implied, of B. Schauerte.
+
+  m=size(XJ,1); % number of samples of p
+  p=size(XI,2); % dimension of samples
+
+  assert(p == size(XJ,2)); % equal dimensions
+  assert(size(XI,1) == 1); % pdist requires XI to be a single sample
+
+  d=zeros(m,1); % initialize output array
+
+  for i=1:m
+    for j=1:p
+      m=(XI(1,j) + XJ(i,j)) / 2;
+			if m ~= 0 % if m == 0, then xi and xj are both 0 ... this way we avoid the problem with (xj - m)^2 / m = (0 - 0)^2 / 0 = 0 / 0 = ?
+				d(i,1) = d(i,1) + ((XI(1,j) - m)^2 / m); % XJ is the model! makes it possible to determine each "likelihood" that XI was drawn from each of the models in XJ
+			end
+    end
+  end

external/base/utilities/histogram_distance/chi_square_statistics_fast.cpp

@@ -0,0 +1,101 @@
+/**
+ * Copyright 2009 B. Schauerte. All rights reserved.
+ * 
+ * Redistribution and use in source and binary forms, with or without 
+ * modification, are permitted provided that the following conditions are 
+ * met:
+ * 
+ *    1. Redistributions of source code must retain the above copyright 
+ *       notice, this list of conditions and the following disclaimer.
+ * 
+ *    2. Redistributions in binary form must reproduce the above copyright 
+ *       notice, this list of conditions and the following disclaimer in 
+ *       the documentation and/or other materials provided with the 
+ *       distribution.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY B. SCHAUERTE ''AS IS'' AND ANY EXPRESS OR 
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 
+ * DISCLAIMED. IN NO EVENT SHALL B. SCHAUERTE OR CONTRIBUTORS BE LIABLE 
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR 
+ * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+ * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *  
+ * The views and conclusions contained in the software and documentation
+ * are those of the authors and should not be interpreted as representing 
+ * official policies, either expressed or implied, of B. Schauerte.
+ */
+
+/**
+ * chi_square_statistics_fast
+ *
+ * Fast C/C++ calculation of the chi-square statistics (compatible with pdist).
+ * (cf. "The Earth Movers' Distance as a Metric for Image Retrieval",
+ *       Y. Rubner, C. Tomasi, L.J. Guibas, 2000)
+ *
+ * @author: B. Schauerte
+ * @date:   2009
+ * @url:    http://cvhci.anthropomatik.kit.edu/~bschauer/
+ */
+#include "mex.h"
+
+#define SQR(x) ((x)*(x))
+
+void
+mexFunction (int nlhs, mxArray* plhs[], 
+	int nrhs, const mxArray* prhs[])
+{
+  mwSize i = 0, j = 0; /* variables for for-loops */
+
+  /* Check number of input parameters */
+	if (nrhs != 2) 
+  {
+  	mexErrMsgTxt("Two inputs required.");
+  } 
+  else 
+		if (nlhs > 1) 
+	  {
+  		mexErrMsgTxt("Wrong number of output arguments.");
+	  }    
+
+  /* Check type of input parameters */
+	if (!mxIsDouble(prhs[0]) || !mxIsDouble(prhs[1])) 
+		mexErrMsgTxt("Input should be double.\n");
+
+	/* Input data */
+	const mxArray* XI = prhs[0];
+	const mxArray* XJ = prhs[1];
+	const double* XI_data = mxGetPr(XI);
+	const double* XJ_data = mxGetPr(XJ);
+	/* some helper variables */
+	const mwSize m = mxGetM(XJ); /* number of samples of p */
+  const mwSize p = mxGetN(XI); /* dimension of samples */
+	if (p != mxGetN(XJ)) 
+		mexErrMsgTxt("Dimension mismatch (1).\n");
+	if (1 != mxGetM(XI)) 
+		mexErrMsgTxt("Dimension mismatch. XI has to be an (1,n) vector.\n");
+	/* Output data */
+	mxArray* OUT = mxCreateNumericMatrix (m, 1, mxDOUBLE_CLASS, mxREAL);
+	plhs[0] = OUT;
+	double* out_data = mxGetPr(OUT);
+
+	for (i = 0; i < m; i++) /* initialize output array */
+		out_data[i] = 0;
+
+	for (j = 0; j < p; j++)
+	{
+		const double xi = XI_data[j];
+		for (i = 0; i < m; i++)
+		{
+			const double mean = (xi + *XJ_data++) / 2.0;
+			if (mean != 0)
+				out_data[i] += SQR(xi - mean) / mean;
+		}
+	}
+
+	return;
+}

external/base/utilities/histogram_distance/hist_dist_example.m

@@ -0,0 +1,143 @@
+% An example of how to use the histogram distance functions for image
+% matching.
+%
+% Please note that this is a demo to show case the usage of the histogram
+% functions. But, in general, matching images solely based on their color
+% histograms ist - imho - not the best idea, unless you have a really large
+% image database.
+%
+% Some of the histogram distance functions have been used for outlier
+% reduction when learning color term/name models from web images, see:
+%
+% [1] B. Schauerte, G. A. Fink, "Web-based Learning of Naturalized Color 
+%     Models for Human-Machine Interaction". In Proceedings of the 12th 
+%     International Conference on Digital Image Computing: Techniques and 
+%     Applications (DICTA), IEEE, Sydney, Australia, December 1-3, 2010. 
+% [2] B. Schauerte, R. Stiefelhagen, "Learning Robust Color Name Models 
+%     from Web Images". In Proceedings of the 21st International Conference
+%     on Pattern Recognition (ICPR), Tsukuba, Japan, November 11-15, 2012
+%
+% If you use and like this code, you are kindly requested to cite some of 
+% the work above. 
+%
+% Anyway, I hope it saves you some work. Have fun with it ;)
+%
+% @author: B. Schauerte
+% @date:   2012,2013
+% @url:    http://cvhci.anthropomatik.kit.edu/~bschauer/
+
+%%
+% Build the .cpp files, if necessary
+if ~exist('chi_square_statistics_fast','file') && exist('./build.m')
+  build;
+end
+
+% Download some random sample images from the Google-512 dataset. For 
+% information about the dataset see: 
+%
+% [1] B. Schauerte, G. A. Fink, "Web-based Learning of Naturalized Color 
+%     Models for Human-Machine Interaction". In Proceedings of the 12th 
+%     International Conference on Digital Image Computing: Techniques and 
+%     Applications (DICTA), IEEE, Sydney, Australia, December 1-3, 2010. 
+% [2] B. Schauerte, R. Stiefelhagen, "Learning Robust Color Name Models 
+%     from Web Images". In Proceedings of the 21st International Conference
+%     on Pattern Recognition (ICPR), Tsukuba, Japan, November 11-15, 2012
+
+%colornames={'red','green','blue','yellow', ...
+%  'pink','purple','brown','orange', ...
+%  'black','grey','white'};
+colornames={'red','green','blue','yellow'};
+fendings={'jpeg','png','gif'};
+tmp_foldername='google-512-samples';
+n_samples = 100;
+
+% download the images in a temporary folder
+if ~exist(tmp_foldername,'dir'), mkdir(tmp_foldername); end % create temporary directory
+filenames=cell(n_samples,1);
+for i=1:n_samples
+  colorname=colornames{randi(numel(colornames))};
+  %colorname=colornames{1};
+  for j=1:numel(fendings)
+    url=sprintf('https://cvhci.anthropomatik.kit.edu/~bschauer/datasets/google-512/images-resized-128/%s+color/%d.%s',colorname,i,fendings{j});
+    filename=sprintf('%s_%d.%s',colorname,i,fendings{j});
+    [~,status] = urlwrite(url,fullfile(tmp_foldername,filename));
+    if status
+      filenames{i} = filename;
+      break; 
+    end
+  end
+end
+
+%%
+% We simply use all files that have already been downloaded
+filenames=dir(fullfile(tmp_foldername,'*_*.*'));
+filenames={filenames.name};
+n_samples=numel(filenames);
+
+%%
+% calculate color image histograms
+n_bins=4;
+edges=(0:(n_bins-1))/n_bins;
+histograms=zeros(n_samples,n_bins*n_bins*n_bins);
+for i=1:n_samples
+  I=imread(fullfile(tmp_foldername,filenames{i}));
+  IR=imresize(I,[64 64]);
+  IR=im2double(IR);
+
+  [~,r_bins] = histc(reshape(IR(:,:,1),1,[]),edges); r_bins = r_bins + 1;
+  [~,g_bins] = histc(reshape(IR(:,:,1),1,[]),edges); g_bins = g_bins + 1;
+  [~,b_bins] = histc(reshape(IR(:,:,1),1,[]),edges); b_bins = b_bins + 1;
+
+  histogram=zeros(n_bins,n_bins,n_bins);
+  for j=1:numel(r_bins)
+    histogram(r_bins(j),g_bins(j),b_bins(j)) = histogram(r_bins(j),g_bins(j),b_bins(j)) + 1;
+  end
+  histograms(i,:) = reshape(histogram,1,[]) / sum(histogram(:)); % normalize, better for all probabilistic methods
+end
+
+%%
+% match histograms and show best matching pairs
+dist_func=@chi_square_statistics_fast;
+% 1. You can use pdist to calculate the distances, iff the distance measure
+%    is symmetric
+%D=squareform(pdist(histograms,dist_func)); % use pdist to calculate the distance for all image pairs 
+% 2. Use the following loop to calculate the distances, iff the measure is
+%    not symmetric
+% D=zeros(size(histograms,1),size(histograms,1));
+% for i=1:size(histograms,1)
+%   for j=1:size(histograms,1)
+%     D(i,j) = dist_func(histograms(i,:),histograms(j,:));
+%   end
+% end
+% 2. ... alternatively, use pdist2
+D=pdist2(histograms,histograms,dist_func);
+
+D(D == 0) = NaN;
+n_show_samples=5; % number of samples for the illustration
+figure('name','Random images (left) with their best (middle) and worst (right) match');
+c = 1;
+rand_indices=randperm(numel(filenames));
+for i=1:n_show_samples
+  % image we want to match
+  I=imread(fullfile(tmp_foldername,filenames{rand_indices(i)}));
+  if numel(size(I)) > 3, I=I(:,:,1:3); end
+  subplot(n_show_samples,3,c); imshow(I); c = c + 1;
+
+  % best match
+  %[d,j]=min(D(rand_indices(i),:)); % if distances are not symmetric, then
+  % it might be useful to try the other order, see below, depending on the
+  % definition of the metric
+  [d,j]=min(D(:,rand_indices(i)));
+  I=imread(fullfile(tmp_foldername,filenames{j}));
+  if numel(size(I)) > 3, I=I(:,:,1:3); end
+  subplot(n_show_samples,3,c); imshow(I); title(sprintf('Dist: %.3f',d*100)); c = c + 1;
+
+  % worst match
+  %[d,j]=max(D(rand_indices(i),:)); % if distances are not symmetric, then
+  % it might be useful to try the other order, see below, depending on the
+  % definition of the metric
+  [d,j]=max(D(:,rand_indices(i)));
+  I=imread(fullfile(tmp_foldername,filenames{j}));
+  if numel(size(I)) > 3, I=I(:,:,1:3); end
+  subplot(n_show_samples,3,c); imshow(I); title(sprintf('Dist: %.3f',d*100)); c = c + 1;
+end

external/base/utilities/histogram_distance/histogram_intersection.m

@@ -0,0 +1,51 @@
+function d=histogram_intersection(XI,XJ)
+  % Implementation of the histogram intersection distance to use with pdist
+  % (cf. "The Earth Movers' Distance as a Metric for Image Retrieval",
+  %      Y. Rubner, C. Tomasi, L.J. Guibas, 2000)
+  %
+  % @author: B. Schauerte
+  % @date:   2009
+  % @url:    http://cvhci.anthropomatik.kit.edu/~bschauer/
+
+  % Copyright 2009 B. Schauerte. All rights reserved.
+  % 
+  % Redistribution and use in source and binary forms, with or without 
+  % modification, are permitted provided that the following conditions are 
+  % met:
+  % 
+  %    1. Redistributions of source code must retain the above copyright 
+  %       notice, this list of conditions and the following disclaimer.
+  % 
+  %    2. Redistributions in binary form must reproduce the above copyright 
+  %       notice, this list of conditions and the following disclaimer in 
+  %       the documentation and/or other materials provided with the 
+  %       distribution.
+  % 
+  % THIS SOFTWARE IS PROVIDED BY B. SCHAUERTE ''AS IS'' AND ANY EXPRESS OR 
+  % IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 
+  % WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 
+  % DISCLAIMED. IN NO EVENT SHALL B. SCHAUERTE OR CONTRIBUTORS BE LIABLE 
+  % FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
+  % CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
+  % SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR 
+  % BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 
+  % WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+  % OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+  % ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+  % 
+  % The views and conclusions contained in the software and documentation
+  % are those of the authors and should not be interpreted as representing 
+  % official policies, either expressed or implied, of B. Schauerte.
+
+  m=size(XJ,1); % number of samples of p
+  p=size(XI,2); % dimension of samples
+
+  assert(p == size(XJ,2)); % equal dimensions
+  assert(size(XI,1) == 1); % pdist requires XI to be a single sample
+
+  d=zeros(m,1); % initialize output array
+
+  sxi=sum(XI);
+  for i=1:m
+    d(i,1) = 1 - (sum(min(XI, XJ(i,:))) / sxi);
+  end