function permutedFeatureVectors = makeAppendedFeatures(rownumber, image, scales, returnall, imageFeatureVector)
% Version 0.1, Feb 2006.
% This code is copyrighted by Ashutosh Saxena, Stanford University
% It is available for non-commercial use only. For commericial use, please
% contact the authors.
% Any publication or report resulting from Use of this code, should cite:
% Learning Depth from Single Monocular Images, Ashutosh Saxena, Sung H. Chung, Andrew Y. Ng, NIPS 2005.
% 3-D Depth Reconstruction from a Single Still Image, Ashutosh Saxena, Sung H. Chung, Andrew Y. Ng, To appear in IJCV 2007.
% For more info and dataset, visit: http://ai.stanford.edu/~asaxena/learningdepth/
% For a better version of the algorithm, also visit: 
% http://ai.stanford.edu/~asaxena/reconstruction3d/



global nGridRow nGridCol featureDirectory
global nStatistics

configurationFile;

if nargin < 4
    returnall = 0;
    imageFeatureVector = 0;
end

numStatistics = nStatistics;
% Per scale
numfeatures = 17 * numStatistics;
% Scales must follow the order of scales in the mat file
numscales = length(scales);

load([featureDirectory 'allTrainingFeatureData_row' num2str(rownumber) ...
	'_' num2str(image) '.mat']);
currowFeatureVector = rowFeatureVector;
numimages = size(currowFeatureVector, 4);
numcolumns = size(currowFeatureVector, 2);

numNonScales = 0;
for i=1:numscales
	if scales(i) < 0
		numNonScales = numNonScales + 1;
	end
end

numNeighbors = 5;
if returnall == 1
	numNeighbors = 7;
end

% Middle Top, bot, left, right order
% The whole image feature is the last features
%permutedFeatureVectors = zeros(1, numcolumns, ...
%	(5*(numscales-numNonScales)+numNonScales)*numfeatures, numimages);
permutedFeatureVectors = zeros(1, numcolumns, ...
	(numNeighbors*(numscales-numNonScales))*numfeatures + numNonScales*68, ...
	numimages);

% Let's load all the necessary feature vectors
for z=1:numscales
	curscale = scales(z);
	i = 0;
	if curscale == 1
		i = 1;
	elseif curscale == 2
		i = 2;
	elseif curscale == 5
		i = 3;
	elseif curscale == -1
		i = 4;
	end
	if curscale > 0
		offset = (curscale - 1)*2 + 1;
		toprow = rownumber - offset;
		botrow = rownumber + offset;
		toptoprow = toprow - offset;
		botbotrow = botrow + offset;
		if toprow < 1
			toprow = 1;
		elseif botrow > nGridRow
			botrow = nGridRow;
		end
		if toptoprow < 1
			toptoprow = 1;
		elseif botbotrow > nGridRow
			botbotrow = nGridRow;
		end

		% The feature at the center
%		permutedFeatureVectors(1, :, ...
%			(5*(z-1)*numfeatures+1):((5*(z-1)+1)*numfeatures), :) = ...
%			currowFeatureVector(1, :, ((i-1)*numfeatures+1):(i*numfeatures), :);
		permutedFeatureVectors(1, :, ...
			(numNeighbors*(z-1)*numfeatures+1):((numNeighbors*(z-1)+1)*numfeatures), :) = ...
			currowFeatureVector(1, :, ((i-1)*numfeatures+1):(i*numfeatures), :);

		neighborid = 1;
		% The feature at the top top
		if returnall == 1
			load([featureDirectory 'allTrainingFeatureData_row' num2str(toptoprow) ...
				'_' num2str(image) '.mat']);
			permutedFeatureVectors(1, :, ...
				((numNeighbors*(z-1)+neighborid)*numfeatures+1):((numNeighbors*(z-1)+neighborid+1)*numfeatures), :) = ...
				rowFeatureVector(1, :, ((i-1)*numfeatures+1):(i*numfeatures), :);
			neighborid = neighborid + 1;
		end

		% The feature to the top
		load([featureDirectory 'allTrainingFeatureData_row' num2str(toprow) ...
			'_' num2str(image) '.mat']);
		permutedFeatureVectors(1, :, ...
			((numNeighbors*(z-1)+neighborid)*numfeatures+1):((numNeighbors*(z-1)+neighborid+1)*numfeatures), :) = ...
			rowFeatureVector(1, :, ((i-1)*numfeatures+1):(i*numfeatures), :);
		neighborid = neighborid + 1;

		% The feature to the bottom bottom
		if returnall == 1
			load([featureDirectory 'allTrainingFeatureData_row' num2str(botbotrow) ...
				'_' num2str(image) '.mat']);
			permutedFeatureVectors(1, :, ...
				((numNeighbors*(z-1)+neighborid)*numfeatures+1):((numNeighbors*(z-1)+neighborid+1)*numfeatures), :) = ...
				rowFeatureVector(1, :, ((i-1)*numfeatures+1):(i*numfeatures), :);
			neighborid = neighborid + 1;
		end

		% The feature to the bottom
		load([featureDirectory 'allTrainingFeatureData_row' num2str(botrow) ...
			'_' num2str(image) '.mat']);
		permutedFeatureVectors(1, :, ...
			((numNeighbors*(z-1)+neighborid)*numfeatures+1):((numNeighbors*(z-1)+neighborid+1)*numfeatures), :) = ...
			rowFeatureVector(1, :, ((i-1)*numfeatures+1):(i*numfeatures), :);
		neighborid = neighborid + 1;

		% The feature to the left
		perm = zeros(1, numcolumns);
		for j=1:numcolumns
			if j <= offset
				perm(1, j) = 1;
			else
				perm(1, j) = j - offset;
			end
		end
		permutedFeatureVectors(1, :, ...
			((numNeighbors*(z-1)+neighborid)*numfeatures+1):((numNeighbors*(z-1)+neighborid+1)*numfeatures), :) = ...
			currowFeatureVector(1, perm, ...
			((i-1)*numfeatures+1):(i*numfeatures), :);
		neighborid = neighborid + 1;

		% The feature to the right
		perm = zeros(1, numcolumns);
		for j=1:numcolumns
			if (j + offset) > numcolumns
				perm(1, j) = numcolumns;
			else
				perm(1, j) = j + offset;
			end
		end
		permutedFeatureVectors(1, :, ...
			((numNeighbors*(z-1)+neighborid)*numfeatures+1):((numNeighbors*(z-1)+neighborid+1)*numfeatures), :) = ...
			currowFeatureVector(1, perm, ...
			((i-1)*numfeatures+1):(i*numfeatures), :);
		neighborid = neighborid + 1;
	elseif curscale == -1
		permutedFeatureVectors(1, :, ...
			(numNeighbors*(z-1)*numfeatures+1):(numNeighbors*(z-1)*numfeatures+68), :) = ...
			currowFeatureVector(1, :, ((i-1)*numfeatures+1):end, :);
	elseif curscale == -2
		% Let's load the image data
		if nargin < 5
			load([featureDirectory 'allTrainingFeatureData_image' num2str(image) '.mat']);
		end
		for cols=1:nGridCol
			permutedFeatureVectors(1, cols, ...
				(numNeighbors*(z-2)*numfeatures+68+1):(numNeighbors*(z-2)*numfeatures+68+68), :) = ...
				imageFeatureVector(1, 1, :, :);
		end
	end
end