Creating an image datastore from the images already loaded in the memory

I was working with the [The Street View House Numbers (SVHN) Dataset][1] and primarily with the `train_32x32.mat` file given. Loading the file as:

load train_32x32.mat

loads 2 variables `X` and `y` in the workspace.

The dimensions of each `X` and `y` are `32 x 32 x 3 x 73257` and `73257 x 1` respectively. I am used to creating an `imageDatastore` and then using `augmentedImageDatastore` to put some form of pre-processing before attempting to create a `CNN`.

But now since `X` already contains the image pixels for `73257` images, is there a way I could still create an `imageDatastore`? For what I could make through the [documentation of imageDatstore][2], it only accepts the physical location of the image directory.

[1]: http://ufldl.stanford.edu/housenumbers/

[2]: https://in.mathworks.com/help/matlab/ref/matlab.io.datastore.imagedatastore.html

Answer 1

[Image Analyst]

It doesn't look like there is a way unless you save them to disk files first.

You could then delete those files when your program exits if you want.