The ImageDatastore currently does not support for CNN regression problems.
Please give the input data via table data for specifying image files as follows.
% Define the image files as cell arrays
filename = {'C:\****\img\img1.png';
'C:\****\img\img2.png';
'C:\****\img\img3.png';
'C:\****\img\img4.png'};
% Define the target values
value = [0.1;0.7;0.3;0.5];
% Create a table data
tbl = table(filename, value);
% Data Augmentation
imageAugmenter = imageDataAugmenter( ...
'RandRotation',[-20,20], ...
'RandXTranslation',[-3 3], ...
'RandYTranslation',[-3 3])
augimds = augmentedImageDatastore([128,128, 3],tbl,'DataAugmentation',imageAugmenter);
% Train network for CNN regression problem
net = trainNetwork(augimds, layers, options)
