How can i input a 2-d extracted features by wavelet scattering to a deep learning network

Question

Nagwa megahed 2022-3-4

回答： Shubham 2024-1-23

dear all,

please i ask to a way that enable me to add a layer that hold the extracted features from wavelet scattering to a deep network?

the input in the deep network usually an input layer that hold the image datastore

yanqi liu 2022-3-7

may be use feature as input 4-D matrix，and use Layers to train it

if possible，may be upload your data to analysis

Answer 1

Shubham 2024-1-23

Hi Nagwa,

To input 2-D extracted features by wavelet scattering into a deep learning network, you can follow these general steps:

Extract Wavelet Scattering Features: Use the Wavelet Scattering Transform to extract features from your images. In MATLAB, you can use the waveletScattering object to perform this operation. The output will be a set of coefficients that represent your image in the scattering domain. Refer to this documentation link: https://in.mathworks.com/help/wavelet/ref/waveletscattering.html
Prepare the Data: Organize the scattering coefficients in a format suitable for training. If you have multiple images, you should have a corresponding set of scattering coefficients for each image. These coefficients can be arranged into a matrix or a cell array, depending on their size and consistency.
Define the Input Layer: Instead of using an imageInputLayer, you will use a featureInputLayer with the size corresponding to the number of scattering coefficients for each image. This will be the first layer of your deep learning network.
Build the Deep Learning Network: Define the rest of the deep learning network architecture using layers like fullyConnectedLayer, reluLayer, softmaxLayer, and classificationLayer. You can also include other types of layers depending on your specific application and network design.
Train the Network: Use the trainNetwork function to train the network. You will need to provide the scattering coefficients as the input data and the corresponding labels as the target data.