Data format in neural network

Hi all, Kindly i really appreciate your help , i have a plan to make skin color segmentation by using neural network , i just have a problem how to format my database ,input vectors it's gona be YCbCr color space and the target 1,0 skin,non skin

Thanks

Answer 1

[Greg Heath]

N pairs of I-dimensional inputs and 1-dimensional targets.

 [ I N ] = size(inputs)

 [ O N ] = size(targets) % O =1

 net = patternnet;   % First consider defaults

 help patternnet

If you have an older version of the NNTBX, try newpr.

 help newpr

You may want to practice on the classification/pattern-recognition data in

 help nndatasets.

Thank you for formally accepting my answer

Greg

Answer 2

[Yasir]

Hi Greg, i really appreciate your kind reply but still now not so clear let's suppose that RGB (119,79,53) these pixels represent skin color while (200,34,23) nonskin are am going to format it as below

[119 79 53;200 34 23] %input vectors [1 1 1 ;0 0 0]%output vector ------ 1- skin , 0-nonskin

and then i fed up these to NN to train it , i really appreciate your clarification because the examples in MATLAB for wine classification the DB not so clear just [X,T]=wine_dataset; % X input ,T target

Thanks for your cooperation

[Greg Heath]

PLEASE USE THE COMMENT BOX, NOT THE ANSWER BOX FOR FOLLOW UP QUESTIONS AND COMMENTS.

NOT ONLY IS IT CONFUSING, IT MESSES UP MATLABS QUESTION/ANSWER/COMMENT BEAN COUNTING.

THANKS.

GREG

[Greg Heath]

I don't understand your confusion. If an I dimensional input column vector corresponds to an O dimensional target output column vector, the dimensions of the input and output target matrices are

 [ I N ] = size(input)
 [ O N ] = size(target)

 [ X, T ] = wine_dataset;
 [ I N ] = size(X)
 [ O N ] = size(T)
 whos X T

 I =    13
 N =   178
 O =     3
 N =   178

Name       Size             Bytes  Class     Attributes

T          3x178             4272  double              
X         13x178            18512  double

Hope this helps.

Greg

[Yasir]

apology okay i will try to overcome it Thanks a lot Greg

[Image Analyst]

I marked Greg's answer as accepted for you because you forgot to.