image input for trained neural network

I'm very new to Neural networks in Matlab and i have this trained neural network

1   'imageinput'    Image Input             16x16x1 images with 'zerocenter' normalization
2   'relu_1'        ReLU                    ReLU
3   'fc_1'          Fully Connected         16 fully connected layer
4   'relu_2'        ReLU                    ReLU
5   'fc_2'          Fully Connected         10 fully connected layer
6   'softmax'       Softmax                 softmax
7   'classoutput'   Classification Output   crossentropyex with '0', '1', and 8 other classes

I have 20000 16*16 binary images for input and validation which goes fine but then i want to classify a random image with the trained network (the size of the input image is 16 * 16) but i get

>>> net(image)
Index exceeds matrix dimensions.

Using R2017a

Answer 1

[Parag]

Hi, the issue may occur because MATLAB's “trainNetwork” function expects the input image to be in a 4D format: (Height, Width, Channels, Batch Size). Since your trained network was designed with an image input layer (imageInputLayer([16 16 1])), it requires the input image to match this format.

Please refer to this code for implementation on dummy data

% Define Layers 
layers = [ 
    imageInputLayer([16 16 1])   
    convolution2dLayer(3, 8, 'Padding', 1) % Use numeric padding instead of 'same' 
    reluLayer 
    maxPooling2dLayer(2, 'Stride', 2) 
    fullyConnectedLayer(10)   
    softmaxLayer 
    classificationLayer]; 

% Define Training Options  

options = trainingOptions("sgdm", ... 
    "MaxEpochs", 5, ... 
    "MiniBatchSize", 32, ... 
    "Verbose", false);  
% Generate Random Binary Training Data (1000 samples) 
numTrain = 1000; 
trainImages = randi([0, 1], 16, 16, 1, numTrain); % 4D array (Height, Width, Channels, Batch) 
trainLabels = categorical(randi([0, 9], numTrain, 1)); % Labels from 0 to 9 
% Train Network 
net = trainNetwork(trainImages, trainLabels, layers, options); 
% Test with a Random 16x16 Image 
testImage = randi([0, 1], 16, 16, 1); % Binary test image 
testImage = reshape(testImage, [16 16 1 1]); % Ensure 4D input format 
predLabel = classify(net, testImage); 
disp("Predicted Label: " + string(predLabel)) 

Predicted Label: 6