Viewing the ouput of a neural network

Hello,

I have tried out the matlab example on 'Training a deep neural network for image classification'. I am getting the results as in the specified example. But my question is how do I view an output for a given input ? Note : My commands are exactly similar to the example just for trial purposes.

So after deepnet is formed : I do something like :

   y = deepnet(xTest);
  ----------------------------------Till here same as the code
  finalnet = configure(deepnet,xTrainImages{1}, tTrain(:,1))
  U = imread('YYY.png')
  imshow(U)
  images_chk{1} = U
  idj = sim(finalnet, images_chk{1})

This results in a weird decimal matrix, I would ideally like it to output the class of the input image eg a column containing 1 for 1 entry and 0 for remaining. How can I do that ?

Thanks

Answer 1

[Greg Heath]

For classification into c classes, the target matrix columns should be {0,1} columns from the eye(c) matrix. The transformation between target (and output) vectors and class indices are provided by the VEC2IND & IND2VEC functions :

>> clc, classind = [ 5 3 1 4 2  ], 
   target = full(ind2vec(classind)),  
   output = target + 0.1*randn(5),
   classind = vec2ind(output)

 classind =   5     3     1     4     2

 target =     0     0     1     0     0
              0     0     0     0     1
              0     1     0     0     0
              0     0     0     1     0
              1     0     0     0     0

output =

    -0.053682      0.13094        1.073     0.057452     0.079318
    -0.030203     -0.10447     0.049075     0.028184      0.91016
      0.18136      0.96517    -0.058613      0.11393     0.015624
     0.091485      0.14126      0.07449      0.95741      0.15973
      0.99429      0.15024    -0.082815     0.063614     0.011244

 classind =      5     3     1     4     2

Hope this helps

Thank you for formally accepting my answer

Greg

Answer 2

[Walter Roberson]

It is common that you would need to round() to get the class number.