cnn for feature extraction

i am doing project on image classification.i have converted image in R,G and B channels.now i need to extract features from each channel using cnn.how can i use cnn for feature extraction in image.

Answer 1

[Sourav Bairagya]

As you have your RGB images ready, then you can define your custom convolutional neural network using 'dlNetwork' object and train it to extract features out of it.

For training you can leverage this link:

https://www.mathworks.com/help/deeplearning/ref/dlnetwork.html

For extracting feature from the trained network, use 'predict' function. To know more about this leverage this link:

https://www.mathworks.com/help/deeplearning/ref/predict.html?s_tid=doc_ta

Answer 2

[esther MUKOYA]

Kindly could you share on the method used to convert the images to RGB?

Answer 3

[Shahram Taheri]

Hi,

imds = imageDatastore('Your Dataset PATH', ...
    'IncludeSubfolders',true,'LabelSource','foldernames');
imds = shuffle(imds);
[imdsTrain,imdsValidation] = splitEachLabel(imds,0.7);
net = resnet18;
%net=densenet201;
inputSize = net.Layers(1).InputSize;
%analyzeNetwork(net)

augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain,'ColorPreprocessing','gray2rgb');
augimdsTest = augmentedImageDatastore(inputSize(1:2),imdsValidation,'ColorPreprocessing','gray2rgb');
layer = 'pool5';
featuresTrain = activations(net,augimdsTrain,layer,'OutputAs','rows');
featuresTest = activations(net,augimdsTest,layer,'OutputAs','rows');
YTrain = imdsTrain.Labels;
YTest = imdsValidation.Labels;
classifier = fitcecoc(featuresTrain,YTrain);
YPred = predict(classifier,featuresTest);