Convolution Neural Network (CNN) accommodates high dimension features and large amount of data with high computation. In this paper, a CNN model for face recognition system that is based upon random forest learning approach is presented. It extracts the convolution neural network based
linear and non-linear features of images. Random forest learns the linear and nonlinear features with different number of trees. The random forest learning is used with adaptive boosting algorithm for enhancing the recognition accuracy. It selects effective tree by boosting approach using
adaptive threshold at testing time. For performance evaluation, the proposed boosted random forest based CNN model is compared with the existing model of soft-max learner based CNN model. The YALE dataset is used that contains the images of 38 persons, having 64 images for each person. The
proposed approach achieves significant accuracy of 99.7%.
A new Rule to Constrain Convolution Neural Network Architecture in Face Recognition System