अमूर्त

FRONTAL FACE DETECTION METHODS –NEURAL NETWORKS AND AGGRESSIVE LEARNING ALGORITHM

Sushma Jaiswal, Dr. Sarita Singh Bhadauria, Dr.Rakesh Singh Jadon

In this Case Study & report, a face detection method is presented. Face detection is the first step of face Recognition methods. Face detection is a difficult task in Pattern. There are different methods of face detection namely-Knowledge Based Face Detection Methods, Feature Based Face Detection Methods, Template Based Face Detection Methods and Appearnce Based Face Detection Methods. But here we divided basically in two methods for face detection (i) image based methods (ii) feature based methods. We have developed an intermediate system, using a boosting algorithm to train a classifier which is capable of processing images rapidly while having high detection rates. AdaBoost is a kind of large margin classifiers and is efficient for on-line learning. In order to adapt the AdaBoost algorithm to fast face recognition, the original Adaboost which uses all given features is compared with the boosting along feature dimensions. The comparable results assure the use of the latter, which is faster for classification. The main idea in the building of the detector is a learning algorithm based on boosting: AdaBoost. AdaBoost is an aggressive learning algorithm which produces a strong classifier by choosing visual features in a family of simple classifiers and combining them linearly. The family of simple classifiers contains simple rectangular wavelets which are reminiscent of the Haar basis. Their simplicity and a new image representation called Integral Image allow a very quick computing of these Haarlike features. Then a structure in cascade is introduced in order to reject quickly the easy to classify background regions and focus on the harder to classify windows. For this, classifiers with an increasingly complexity are combined sequentially. This improves both, the detection speed and the detection efficiency. The detection of faces in input images is proceeded using a scanning window at different scales which permits to detect faces of every size without resampling the original image. On the other hand, the structure of the final classifier allows a realtime implementation of the detector. Due to some limitation of neural network based methods we adopt the Adaboost algorithm for face detection. Here we present some results on real world examples are presented. Our detector found good detection rates with frontal faces and the method can be easily adapted to other object detection tasks by changing the contents of the training dataset.