How to give target data in nntool or in nnstart?

I have created dataset of 94 B-scan images (76 training + 18 testing) consists of 3 classes (metal pipe, steel box, plastic box). I have extracted statistical features from images the size of training features is 76*6. and I have assigned labels using traindb.Labels, I can able to load input data as training features from work space but Iam unable to load target data.

My goal is to classify the three objects using NN classifier. I have classified these objects using ECOC-SVM but I'm unable to load target data using nntool. I'm attaching my code and database for your reference. please help me.

[santhosh kumar buddepu]

my labels are 76*1 categorical (metal pipe-19, plastic box-28, steel box-29). I need to classify into 3 classes and my target should be 76*3 which is in the form of

1 1 1....1(19) 0 0 0... (remaining)

0 0 0..0(19) 1 1 1...1(28) 0 0...0(remaining)

0 0 0......0(47) 1 1 1...1(29)

how to make target like this? please help me

Answer 1

[Samay Sagar]

To load target data for use with a Neural Network classifier, you need to convert your categorical labels into a format that the neural network can understand. This usually means transforming the labels into a one-hot encoded format if you are using a network that outputs class probabilities for each class.

Here is how you can modify your labels:

% Assuming traininglabels is a 76x1 categorical array with categories:
% 'metal pipe', 'plastic box', 'steel box'

% Initialize the target matrix with zeros
numImages = numel(traininglabels); % Number of training images
categoriesList = categories(traininglabels); % Get the list of categories
numClasses = numel(categoriesList); % Number of unique classes
targetMatrix = zeros(numImages, numClasses);

% For each class, set the appropriate entries in the target matrix to 1
for classIndex = 1:numClasses
    className = categoriesList{classIndex}; % Access the category name correctly
    % Find all rows in traininglabels that belong to the current class
    rowsOfClass = traininglabels == className;
    % Set the corresponding entries in the target matrix to 1
    targetMatrix(rowsOfClass, classIndex) = 1;
end

% Now, targetMatrix is your desired 76x3 target matrix

You can follow a similar approach to prepare your testing target data.

Hope this helps!