Skip to content

Commit

Permalink
Initial Commit
Browse files Browse the repository at this point in the history
  • Loading branch information
@javixw
javixw authored Jun 2, 2020
1 parent 1772265 commit 2af6e6d
Show file tree
Hide file tree
Showing 9 changed files with 587 additions and 0 deletions.
10 changes: 10 additions & 0 deletions add_path.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
addpath('CoordsTransfrom');
addpath('icosahedron2sphere');
addpath('Projection');
addpath(genpath('BasicProcessing'));
addpath('visualization');
addpath('definitiva_ajustado');
addpath('imagenesGT');
addpath('ImagenesResultadosDef');
addpath('workspace');
addpath('imgs');
258 changes: 258 additions & 0 deletions agregation.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,258 @@
function [matrix_label_end_agregation_1, matrix_score_end_agregation_1, matrix_label_end_agregation_2, matrix_score_end_agregation_2, matrix_label_end_agregation_3, matrix_score_end_agregation_3] = agregation(nlabels, nx, ny, bbox_600_300)
label_label_def = cell(1,nlabels);
label_score_def = cell(1,nlabels);
label_score_def_ag = cell(1,nlabels);
label_score = cell(1,nlabels);
for ilabels=1:nlabels
label_score{1,ilabels} = vertcat(bbox_600_300{:,ilabels});
if isempty(label_score{1,ilabels}) == 0
[f,s] = size(label_score{1,ilabels});
if ilabels == 1
label_score{1,ilabels} = [label_score{1,ilabels} ((182)*ones(f,1))];
label_score_def{1,ilabels} = accumarray([label_score{1,ilabels}(:,1),label_score{1,ilabels}(:,2)],label_score{1,ilabels}(:,3),[nx ny],@max);
label_score_def_ag{1,ilabels} = accumarray([label_score{1,ilabels}(:,1),label_score{1,ilabels}(:,2)],label_score{1,ilabels}(:,3),[nx ny]);
label_label_def{1,ilabels} = accumarray([label_score{1,ilabels}(:,1),label_score{1,ilabels}(:,2)],label_score{1,ilabels}(:,4),[nx ny], @max);
else
label_score{1,ilabels} = [label_score{1,ilabels} ((ilabels-1)*ones(f,1))];
label_score_def{1,ilabels} = accumarray([label_score{1,ilabels}(:,1),label_score{1,ilabels}(:,2)],label_score{1,ilabels}(:,3),[nx ny],@max);
label_score_def_ag{1,ilabels} = accumarray([label_score{1,ilabels}(:,1),label_score{1,ilabels}(:,2)],label_score{1,ilabels}(:,3),[nx ny]);
label_label_def{1,ilabels} = accumarray([label_score{1,ilabels}(:,1),label_score{1,ilabels}(:,2)],label_score{1,ilabels}(:,4),[nx ny], @max);
end
end
end
clear ilabels iimages bbox_600_300_b bbox_600_300 label_score f s
matrix_per_label_def = cell(1,nlabels);
matrix_score_per_label_def = cell(1,nlabels);
matrix_score_per_label_def_ag = cell(1,nlabels);
for k=1:nlabels
matrix_per_label_def{1,k} = label_label_def{1,k}(:);
matrix_score_per_label_def{1,k} = label_score_def{1,k}(:);
matrix_score_per_label_def_ag{1,k} = label_score_def_ag{1,k}(:);
end
clear label_label_def label_score_def label_score_def_ags
%maximum point average agregation and weighted maximum point average agregation
an_label = [];
an_score = [];
an_label_ag = [];
an_score_ag = [];
count = 0;
an_cell_array = cell(nx,ny);
for pos=1:(nx*ny)
an = [];
an_agregation = [];
for k=1:nlabels
if isempty(matrix_per_label_def{1,k}) == 0
if matrix_per_label_def{1,k}(pos) ~= 0
an = [an; single(pos) (matrix_per_label_def{1,k}(pos)) matrix_score_per_label_def{1,k}(pos)];
an_agregation = [an_agregation; single(pos) (matrix_per_label_def{1,k}(pos)) matrix_score_per_label_def_ag{1,k}(pos)];
end
end
end
if isempty(an) == 0
escore = accumarray(an(:,1),an(:,3),[],@max);
[rowscore,colscore] = find(an==escore(end));
[s,z] = size(rowscore);
if (s > 1)
rowscore = rowscore(1);
colscore = colscore(1);
count = count + 1;
end
an_label = [an_label; an(rowscore,(colscore-1))];
an_score = [an_score; an(rowscore,colscore)];
clear rowscore colscore an
else
an_label = [an_label; NaN];
an_score = [an_score; NaN];
end
if isempty(an_agregation) == 0
an_cell_array{pos} = an_agregation;
[max_sco,index] = max(accumarray(an_agregation(:,2),an_agregation(:,3)));
an_label_ag = [an_label_ag; index];
an_score_ag = [an_score_ag; max_sco];

clear max_sco index an_agregation
else
an_agregation = [an_agregation; single(pos) 200 0];
an_cell_array{pos} = an_agregation;
an_label_ag = [an_label_ag; NaN];
an_score_ag = [an_score_ag; NaN];
clear an_agregation
end
end
matrix_label_end_agregation_1 = reshape(an_label,[nx,ny]);
matrix_score_end_agregation_1 = reshape(an_score,[nx,ny]);
matrix_label_end_agregation_2 = reshape(an_label_ag,[nx,ny]);
matrix_score_end_agregation_2 = reshape(an_score_ag,[nx,ny]);

%weighted maximum local average agregation
an_res_label = zeros(nx,ny);
an_res_score = zeros(nx,ny);
sigma = 0.5;
for i=1:(nx*ny)
[y,x] = ind2sub([nx ny],i);
if (y == 1) && (x == 1)
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(2:3,2:3);
minor_column = x; larger_column = x+1; minor_row = y; larger_row = y+1;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:2
for j=1:2
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (x == 1) && (y == nx)
%y = nx; x = 1;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(1:2,2:3);
minor_column = x; larger_column = x+1; minor_row = y-1; larger_row = y;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:2
for j=1:2
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (x == ny) && (y == 1)
%y = 1; x = ny;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(2:3,1:2);
minor_column = x-1; larger_column = x; minor_row = y; larger_row = y+1;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:2
for j=1:2
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (x == ny) && (y == nx)
%y = nx; x = ny;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(1:2,1:2);
minor_column = x-1; larger_column = x; minor_row = y-1; larger_row = y;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:2
for j=1:2
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (x == 1)
%y = 3; x = 1;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(1:3,2:3);
minor_column = x; larger_column = x+1; minor_row = y-1; larger_row = y+1;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:3
for j=1:2
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (y == 1)
%x = 3; y = 1;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(2:3,1:3);
minor_column = x-1; larger_column = x+1; minor_row = y; larger_row = y+1;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:2
for j=1:3
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (y == nx)
%y = nx; x = 3;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(1:2,1:3);
minor_column = x-1; larger_column = x+1; minor_row = y-1; larger_row = y;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:2
for j=1:3
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
elseif (x == ny)
%y = 2; x = ny;
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
h_new = h(1:3,1:2);
minor_column = x-1; larger_column = x; minor_row = y-1; larger_row = y+1;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:3
for j=1:2
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h_new(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
else
hsize = [3 3];
h = fspecial('gaussian',hsize,sigma);
minor_column = x-1; larger_column = x+1; minor_row = y-1; larger_row = y+1;
auu = an_cell_array(minor_row:larger_row,minor_column:larger_column);
a = [];
for ii=1:3
for j=1:3
auu{ii,j}(:,3) = auu{ii,j}(:,3)*h(ii,j);
a = [a; auu{ii,j}(:,2) auu{ii,j}(:,3)];
end
end
accum_a = accumarray(a(:,1), a(:,2), [], @sum);
[C,I] = max(accum_a);
an_res_label(y,x) = I;
an_res_score(y,x) = C;
end

end
matrix_label_end_agregation_3 = an_res_label;
matrix_score_end_agregation_3 = an_res_score;
clear matrix_per_label_def an_cell_array an_res_label an_res_score matrix_score_per_label_def matrix_score_per_label_def_ag an_label an_score an_label_ag an_score_ag
end
68 changes: 68 additions & 0 deletions cleanup_database.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,68 @@
function [interesting_imagefiles, interesting_images, target_ima_pos]= cleanup_database(target_heading, target_pitch, imagestructure, dir_ima_pre)
target_heading_pitch = [imagestructure.images.heading; imagestructure.images.pitch];
target_ima_pos = find(target_heading_pitch(1,:) == target_heading & target_heading_pitch(2,:) == target_pitch);
if isempty(target_ima_pos) == 1
fprintf('Target heading and pitch not found\n');
else
%heading_obj = imagestructure.images(target_ima_pos).heading;
heading_obj_max = target_heading + 90;
flag_heading_obj_max = 0;
if (heading_obj_max > 360)
heading_obj_max = heading_obj_max - 360;
flag_heading_obj_max = 1;
end
heading_obj_min = target_heading - 90;
flag_heading_obj_min = 0;
if (heading_obj_min < 0)
heading_obj_min = 360 + heading_obj_min ;
flag_heading_obj_min = 1;
end
allCN = [imagestructure.images.heading];
if (isempty(find(allCN == heading_obj_max, 1)) == 1) && (isempty(find(allCN == heading_obj_min, 1)) == 1)
distmin = abs(allCN-heading_obj_min);
distmax = abs(allCN-heading_obj_max);
minDistmin = min(distmin);
minDistmax = min(distmax);
idxmin = find(distmin == minDistmin);
idxmax = find(distmax == minDistmax);
[fidxmin, sidxmin] = size(idxmin);
[fidxmax, sidxmax] = size(idxmax);
idxmin_post = idxmin(1:sidxmin/2);
idxmax_post = idxmax(1+sidxmax/2:sidxmax);
[fidxmin_post, sidxmin_post] = size(idxmin_post);
[fidxmax_post, sidxmax_post] = size(idxmax_post);
if (flag_heading_obj_min == 1) || (flag_heading_obj_max == 1)
interesting_images = [imagestructure.images(1:idxmax_post(end)); imagestructure.images((idxmin_post(1)):(end))];
else
interesting_images = imagestructure.images((idxmin_post(1)):idxmax_post(end));
end
else
distmin = abs(allCN-heading_obj_min);
distmax = abs(allCN-heading_obj_max);
minDistmin = min(distmin);
minDistmax = min(distmax);
idxmin_post = find(distmin == minDistmin);
idxmax_post = find(distmax == minDistmax);
if (flag_heading_obj_min == 1) || (flag_heading_obj_max == 1)
interesting_images = [imagestructure.images(1:idxmax_post(end)); imagestructure.images((idxmin_post(1)):(end))];
else
interesting_images = imagestructure.images((idxmin_post(1)):idxmax_post(end));
end
end
imagefiles = dir(strcat(dir_ima_pre,'/*.jpg'));
if (flag_heading_obj_min == 1) || (flag_heading_obj_max == 1)
interesting_imagefiles = [struct('name', {imagefiles(1:idxmax_post(end)).name}, 'folder', {imagefiles(1:idxmax_post(end)).folder}),struct('name', {imagefiles(idxmin_post(1):end).name}, 'folder', {imagefiles(idxmin_post(1):end).folder})];
else
interesting_imagefiles = struct('name', {imagefiles(((interesting_images(1).seqNumber)+1):((interesting_images(end).seqNumber)+1)).name}, 'folder', {imagefiles(((interesting_images(1).seqNumber)+1):((interesting_images(end).seqNumber)+1)).folder});
end

interesting_imagefiles(1).folder_name = [];
% target_pitch_max = target_pitch + 20;
% target_pitch_min = target_pitch - 20;
% idx = [interesting_images(:).pitch]>target_pitch_max;
% interesting_images = interesting_images(~idx);
% interesting_imagefiles = interesting_imagefiles(~idx);
% idx = [interesting_images(:).pitch]<target_pitch_min;
% interesting_images = interesting_images(~idx);
% interesting_imagefiles = interesting_imagefiles(~idx);
end
28 changes: 28 additions & 0 deletions coordinates.m
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,28 @@
function [class_ima, score_class_ima, interesting_imagefiles] = coordinates(ninteresting_files,interesting_imagefiles, dir_ima_post, nlabels)
sco = cell(1,ninteresting_files);
lab = cell(1,ninteresting_files);
for ii=1:ninteresting_files
interesting_imagefiles(ii).folder_name = double(imread(strcat(interesting_imagefiles(ii).folder,'/',interesting_imagefiles(ii).name)));
lab{1,ii} = uint8(csvread(sprintf(strcat(dir_ima_post,'/labelmaps/%s.csv'),erase(interesting_imagefiles(ii).name,'.jpg'))));
sco{1,ii} = single(csvread(sprintf(strcat(dir_ima_post,'/scores/%s.csv'),erase(interesting_imagefiles(ii).name,'.jpg'))));
end

clear imagefiles
[f,s] = size(lab{1,1});
score_class_ima = cell(ninteresting_files,nlabels);
class_ima = cell(ninteresting_files,nlabels);
for iimages=1:ninteresting_files
for ilabels=1:nlabels
for j=1:s
for i=1:f
if lab{1,iimages}(i,j) == (ilabels-1)
class_ima{iimages,ilabels} = uint16([class_ima{iimages,ilabels} ; j i]);
score_class_ima{iimages,ilabels} = single([score_class_ima{iimages,ilabels} ; sco{1,iimages}(i,j)]);
end
end
end
end
end


end
Loading

0 comments on commit 2af6e6d

Please sign in to comment.