Typically, especially for huge sizes, some form of feature extraction is used to reduce the size of the input.
So, instead of image --> image(:), you get image --> featurevector.
If the number of extracted features is I, then the input matrix for N images has size
Next, if there are c categories with Ni (1=1:c) members each, the corresponding target matrix has N columns from the c dimensional unit matrix eye(c) so that
[ c N ] = size(target),
sum(target) = ones(1,N),
sum(target') = [ N1, N2, ...
and
sum(sum(target)) = sum(sum(target')) = sum(target(:)) = N
Hope this helps.
Thank you for formally accepting my answer
Greg
Best Answer