The application of artificial neural network to detect the position of human face in digital image

Abstract EN

One of the important parts in human face recognition is detecting face position.

In this paper a implement face position detection experiment using (Artificial Neural Network, ANN) to give outputs of human face number, position and dimension as found in a digital image.

The system is trained using available ace samples.

Quickprop algorithm and active learning method are used to speed the system training process up.

And also indicate the comparison of the training time with standard Backpropagation algorithm and the training with Quickprop algorithm.

The experiment is conducted using 200, 300 and 400 data.

For each of the trainings, the iteration is stopped when the error value reaches 0.05.

It is observed that the bigger number of the training data of the Quickprop algorithm causes significant increase in the training rate.

Based on the results of the experiment with 14 files containing 273 face images, the face detection system gives 70.24 % detection rate and 62 false positives.