My code runs as a m file but not as an exe. What to do?

The code runs perfectly well in Matlab, however, when I compile an exe file and run that, it gives me the following error:

"subscript indices must either be real positive integers or logicals"

The code is as follows:

face=faceDetect();
dEye=eyeDistance(face);
[dh,dv,A]=mouthDistance(face);
dEyebrow=eyebrowDistance(face);
[f,n]=wrinkles(face);
x=[];
x(1,1)=dEye
x(1,2)=dEyebrow
x(1,3)=dh
x(1,4)=dv
x(1,5)=A
x(1,6)=f
x(1,7)=n
x=transpose(x);
load ('C:\xampp\htdocs\Test\Final\trained_net_3lm.mat');
dlmwrite('C:\xampp\htdocs\Test\testx.txt', x);
y=net(x); %*** HERE ***
dlmwrite('C:\xampp\htdocs\Test\test.txt', y);

The line marked HERE is the one throwing out the error.

Help would be really appreciated. :)

Answer 1

[Image Analyst]

I guess net is one of your arrays. So I'd guess that one of the elements of x is either zero, a negative number, or a fractional value. Type out x before you use it. See the FAQ on this extremely frequently asked question: http://matlab.wikia.com/wiki/FAQ#How_do_I_fix_the_error_.22Subscript_indices_must_either_be_real_positive_integers_or_logicals..22.3F

[AshM]

I think it shouldn't matter what are the contents of the array x as long as the indices are not either negative or zero or floating point number. And even so, the code runs perfectly well in matlab as a .m file but it gives the error mentioned above when ran as an exe file

[Image Analyst]

Please add these lines before the net(x) line and tell us what you see in the console window:

for k = 1 : length(x)
    fprintf('x(%d) = %f\n', k, x(k));
end

And please attach this file: 'C:\xampp\htdocs\Test\testx.txt' with the paper clip icon.

Answer 2

[Steven Lord]

I suspect net is an object created using Neural Network Toolbox functions that you loaded from the MAT-file. If so realize that without some indication MATLAB Compiler can detect via static analysis of your code, it has no way of knowing that it needs to include the functionality from Neural Network Toolbox in your application. You'll need to use the %#function pragma to give it that hint. See in particular the documentation topic linked in Example 3 on that page.

I'd also consider calling load with an output argument rather than just "poofing" the variable into the workspace. Use something like:

mydata = load(...);
y = mydata.net(x);

[AshM]

I tried your suggestion as well. But it is still giving me the same error. What should I do?

[Steven Lord]

Attach the MAT-file C:\xampp\htdocs\Test\Final\trained_net_3lm.mat to this Answer so people can try it out. If that's not possible or you'd prefer not to do that, send the MAT-file along with this question to Technical Support.

[AshM]

This is the trained neural network file. If anybody can try it out and help me figure out this error, I will be grateful.

Answer 3

[Greg Heath]

Nowhere is 'net' defined.

Hope this helps.

Greg

[AshM]

Even after doing that, it doesn't work :(

Answer 4

[Sean de Wolski]

This answer explains it:

How can I compile a pre-trained neural network into a standalone application using the MATLAB Compiler 4.10 (R2009a)? - MATLAB Answers - MATLAB Central (mathworks.com)