Can a cell array be training input for a neural network?

I have created a cell array with aprox 200 letters (capital A,B,C etc ).

Each letter is a matrix, cut from a picture of a car plate. Now my next task is to get those letters into a neural to train. But I have some questions :

a) Is this even possible? Meaning Cell ---> Neural Network b) If not how can I fix it? c) If yes , is it something simple that I am just too blind to see?

Excuse my bad English, I am not native.

[Walter Roberson]

Are the arrays inside the cell array all the same size?

[Vergil234]

Not all of them no. Could you explain please? Trying to understand really.

Answer 1

[TED MOSBY]

Hi,

You can use cell arrays, but you need to convert them into one of the formats which MATLAB's training functions will expect. Thus, you need to convert them into numeric arrays or datastores. You can then simply pass them in your neural network and go ahead with your training.

To know more about "datastore" refer here:

https://www.mathworks.com/help/matlab/ref/datastore.html?searchHighlight=datastores&s_tid=srchtitle_support_results_1_datastores

Here is more information on MATLAB's function "trainnet" used for training deep learning neural network:

https://www.mathworks.com/help/deeplearning/ref/trainnet.html#mw_fe3a9a26-312e-416d-91ba-2c6087a01e45

Hope this helps!

[Walter Roberson]

Note that if you are using a datastore then you will probably want to use https://www.mathworks.com/help/deeplearning/ref/augmentedimagedatastore.html and specify an output size that each image is to be automatically resized to. You can also use options to specify that rgb images are automatically converted to grayscale (or grayscale automatically converted to rgb)