want to dispaly image not found

I am working on face recognisation ,i have train data set and test data set,my code predicts the near face if i select a face image ,but if the selected face is not present in training data set,i want to display a command ."face not found ",can u suggest an method,or code for this

Answer 1

[Walter Roberson]

disp('"face not found"')

I suspect that is not your real question, though. I suspect your real question is "How can I detect that a face is not in the training set?". The answer to that is going to depend upon how your do your face recognition.

I suspect that (one of the) outputs of applying the recognition algorithm is a similarity or probability measure. It will be up to you to decide what value of similarity or probability indicates a non-match. There is no "right" answer in such a matter: there is only tuning it to be more or less generous in deciding that a match has occurred. In some cases you want to err on the side of rejecting matches that are not clearly people you know (e.g., if you were doing access controls that permitted only authorized people), and in other cases you want to err on the side of alerting authorities about a possible but unclear match (e.g., if you are searching for dangerous people or people who might have deliberately disguised themselves.)

[Pat]

what u said is correct "How can I detect that a face is not in the training set?"i am able to detect the near face but it should not be done,i need exact result ,what algorithm must be used for this

[Image Analyst]

What if my face looking at an angle of 0 degrees (straight at the camera) is in the data base, and the "test" image is an image of my face looking at 20 degrees? Or me looking at 0 degrees but with a different facial expression (laughing instead of smiling)? Are those matches or not? Did you call them "near" or not? Would those faces of mine be the same as my face in the training base or not? This project is way more complicated than you might think and is not the kind of thing where someone can give you code in the "Answers" forum.

[Pat]

thank u sir ,can u suggest some method ,or algorithm .(near face sames it detects some other face )

[Walter Roberson]

Perhaps for Pat's requirements, it _is_ relatively simple: perhaps Pat needs to detect *exact* copies of the training images and reject everything else.

In such a case, what I might do is use the neural net to find the closest match, and then compare the original of the closest to the input image.

Or perhaps it is a matter of adjusting the weights to be more rigid in the matching process and of using a net big enough to store exact copies of all of the training images...

[Pat]

thank u walter ,i will try on neural network