Do I have to have a dataset to train a Neural Network or can I just insert the folder location?

Below you can find my code and the 5 attemps made at inserting the images.

close all;
clear;
%%GOAL%%
% get 300 images
% x = 640 x 300
% t = 3 x 300

%% attempt 1 %%
% size(x)
% imgFolder = fullfile('D:\train\chair');
% imgSet = imageSet(imgFolder);

%% attempt 2 %%
% [x,t] = image_dataset;
%  [ I N ]  = size(x)         % [ 640 300 ]
%  [ O N ]  = size(t)         % [ 3 300 ]

%% attempt 3 %%
% imds = imageDatastore('D:\test\cup');
% imgRGB = imds.read; 
 
%% attempt 4 %%
% my_dataset = 'D:\train\chair';
% load(my_dataset);

%% attempt 5 %%
% folder = 'D:\train\chair';
% myImgaes =imageSet(folder);
% x = folder;
% t = 
 
net = patternnet(10);
view(net);
[net,tr] = train(net,x,t);
nntraintool;
plotperform(tr)

testX = x(:,tr.testInd);
testT = t(:,tr.testInd);

testY = net(testX);
testIndices = vec2ind(testY);
plotconfusion(testT,testY)

Answer 1

[Eustace Tan]

What should happen is:

1. Read the first image of size m-by-n. (I'm assuming greyscale. Additional work required if RGB.)
2. Reshape to mn-by-1 vector.
3. Repeat for the other 299 images.
4. Organise data into mn-by-300 matrix.
5. Pair up with target matrix t-by-300.