This paper explores techniques used for feature selection. First, convolution neural network with batch normalization (CNN-BN) is proposed to extract features from an input image. Second, face image retrieval is proposed which is a technique of searching and retrieving images from a large dataset. Lastly, the classification of images is proposed where the images are classified into two categories, namely identity classification and pose classification. The feature extraction technique CNN without normalization layer is compared with CNN with normalization. It is also compared with other feature extraction technique like SIFT. This approach is applicable on wild datasets and achieves better performance on Colorferet, LFW and Multi-PIE dataset.
Volume 11 | 04-Special Issue
Pages: 1051-1063