Dimensions of matrices being concatenated are not consistent.

I have 1*512 cell and i want it to convert in matrix so i am trying to use cell2mat but i am getting this error :

Error using cat
Dimensions of matrices being concatenated are not consistent.
Error in cell2mat (line 75)
          m{n} = cat(2,c{n,:});

i have obtained this cell after using

cellfun(@(x) find(x,1,'first'), cimg,'un',0);

Uploading the mat file for 1*512 as well.

[Adam]

You haven't shown any code that actually calls cell2mat.

[prashant singh]

@Adam just using it after my cellfun.

Answer 1

[Cedric]

You have empty ( see comments below! ) arrays in your cell array (where FIND found nothing), so the dimensions do not match for a CAT (the content of some cells is 1x1 and the content of others is 1x0).

If you can get rid of these empty arrays (and hence reduce the number of elements in the output), you can do something like:

 isEm = cellfun( @isempty, ridx ) ;
 data = cell2mat( ridx(~isEm) ) ;

where data contains the 502 "non-empty values". If you need to keep the size of the original data, you can replace empty arrays with e.g. NaNs:

 ridx(isEm) = {NaN} ;
 data = cell2mat( ridx ) ;

and here data has 512 elements, 10 of which are NaN.

[Guillaume]

Unless, the output of cellfun is 2D (which it shouldn't if the input is 1D), cell2mat will succeed even if find returns empty for some values:

>> cell2mat({1, [], 3, 4})
ans = 
   1   3   4

However

>> cell2mat({1, []; 3, 4})
Error using cat 
Dimensions of matrices being concatenated are not consistent. 
Error in cell2mat (line 86)
m = cat(2,m{:});

Puzzingly,

>> cell2mat({1, [], 3; [], 4, 5; 6 7 []})
ans = 
    1    4   3
    6    7   5

That looks like a bug (testing with matlab Online R2017a, don't have matlab installed on this computer)

[Cedric]

Actually I called empty the 1x0 or 0x1 arrays that FIND outputs, and these are an issue when the dimension of size 0 is not matching the CAT direction:

 >> find( (2:4) == 1 )
 ans =
  1×0 empty double row vector

 >> cell2mat( {1, zeros(1,0), 3, 4} )
 ans =
     1     3     4

 >> cell2mat( {1, zeros(1,0), 3, 4}.' )
 Error using cat
  Dimensions of matrices being concatenated are not consistent.
  Error in cell2mat (line 83)
            m{n} = cat(1,c{:,n});

However, I am still scratching my head about your example! R2016b has the same behavior.

[Guillaume]

Oh, yes. I completely overlooked that matlab has a multitude of empty which are not equal (grrr! complained about that years ago).

In the OP case, simply replacing the cell2mat call by

vertcat(thecellarray{:})

would get rid of the error and possibly produce the desired output (provided that getting rid of the empty entries is what is desired).

I can sort of understand why cell2mat produces the output it does in my last example, but I think it would be much better if it returned an error in this case.

[Cedric]

I think that your example should fail in their test suite:

 >> {1, [], 3; [], 4, 5; 6 7 []}
 ans =
  3×3 cell array
    [1]     []    [3]
     []    [4]    [5]
    [6]    [7]     []

 >> cell2mat( ans )
 ans =
     1     4     3
     6     7     5

unless we invoke gravity, making e.g. the 1 fall down on the 6, as the first dimension is pointing downwards on our screens and there is nothing below the 1 ;-)

[Guillaume]

Agreed, it should fail. Reported as a bug to mathworks.

Answer 2

[Image Analyst]

Try this:

ca = {1, [], 3; [], 4, 5; 6 7 []}
m = [ca{:}]
ca2 = {1, [], [3,4,5]; [], 4, 5; 6 [7,2] []}
m = [ca2{:}]

You'll see

ca =
  3×3 cell array
    [1]     []    [3]
     []    [4]    [5]
    [6]    [7]     []
m =
     1     6     4     7     3     5
ca2 =
  3×3 cell array
  [1]              []    [1×3 double]
   []    [         4]    [         5]
  [6]    [1×2 double]              []
m =
   1     6     4     7     2     3     4     5     5

In ca2, you'll note that all arrays are not all the same size, yet it still works.

[Cedric]

Hi Image Analyst,

This is normal, : is linear indexing (column first) a 2D cell array, {} makes the output a CSL, and [] is concatenating.

The following legitimately doesn't work:

 >> ca3 = {1, [], [3,4,5]; [], 4, 5; 6 [7;2] []}  % [7,2] -> [7;2]
 ca3 =
  3×3 cell array
    [1]              []    [1×3 double]
     []    [         4]    [         5]
    [6]    [2×1 double]              []
 >> m = [ca3{:}]
 Error using horzcat
 Dimensions of matrices being concatenated are not consistent.

In Guillaume's example,

 >> cell2mat({1, [], 3; [], 4, 5; 6 7 []})
 ans =
     1     4     3
     6     7     5

it goes through because there are the same number of non-empty elements in each row/column, when it should not, because the positions of some elements are changed relatively to the other in the output. And if there is a difference in the number of non-empty elements, it doesn't work:

 >> cell2mat({1, [], 3; [], 4, 5; 6 7 8})
 Error using cat
 Dimensions of matrices being concatenated are not consistent.
 Error in cell2mat (line 86)
        m = cat(2,m{:});

As far as I am concerned, it should not work in the first case, as it goes through by chance when it should fail to indicate an invalid operation.