Question

I am doing a face recognition project on face recognition based on neural network.It includes training the system on various images.Does that mean that after training the system on various images it recognize people without matching it from any database?Like in character reading if we train our system with all the character can it recognize a new character without any further training?

Answer 1

Face recognition involves a situation where you want to give the system an image (a newly observed face) and have it compare to a set of previously defined images.

The training will consist of developing a neural network that does a better and better job of doing this job (i.e. matching a new image to a stored image). However, even after training, you still need that set of stored images.

If you’re only interested in one face (e.g. you want a system that can learn to recognize you, but no one else), then I suppose you could train a neural network to recognize that image, and then the image properties would be “built in” to the trained neural network. But that seems both less useful and less efficient than training the system to match against a database, and then providing that database.

Or do you mean that you want the system to see a person, and then later “know” that person if they see them again? If so, again, the efficient way would be to train the system to reduce a face to a set of characteristics, and store them in a database for later matching.