test_learn_filt.m

Standard model object recognition matlab code

function [test_result,test_loc] = test_learn_filt (f,r,s)

% FUNCTION [test_result,test_loc] = test_learn_filt (f,r,s)
% 
% Test if the leraned filter f is valid, by trying to find the
% patch from the input r.  The output test_result will give the
% indices of the features from which the matching patch was found.
% The input arguement s specifies the scale from which f was
% learned.

s_tag = ['s' num2str(s)];

i1 = getfield(f, ['i1_' s_tag]);
i2 = getfield(f, ['i2_' s_tag]);
i3 = getfield(f, ['i3_' s_tag]);
siz = getfield(f, ['size_' s_tag]);

f_val = getfield(f, ['f1_' s_tag]);

x  = getfield(r,s_tag);
for i = 1:15%size(f_val,2) % Loop thru features
  for j = 1:size(f_val,1)
    for m = 1:size(x,1)-siz(1)
      for n = 1:size(x,2)-siz(2)
	x_val(j,m,n) = x(i1(j,i)+m,i2(j,i)+n,i3(j,i));
      end
    end
  end

  % Threshold for finding, in case of roundoff error.
  thresh = eps*1000; 
  is_match = squeeze(sum(abs(x_val-repmat(f_val(:,i),[1 m n]))));
  is_found = find(is_match<thresh);

  if length(is_found) > 0
    if length(is_found)>1
      disp(['Threshold too big for feature ' num2str(i)]);
    end
    test_result(i) = 1;
    [m_tmp,n_tmp]  = ind2sub(size(is_match),is_found(1));
    test_loc(1,i)  = m_tmp;
    test_loc(2,i)  = n_tmp; 
  else
    test_result(i)  = 0;
    test_loc(1:2,i) = 0;
  end
end