Deconstructive human face recognition using deep neural network

Confront reproduction or confront optimization as a biometric, encompasses a few points of interest in measurable application. The estimate and structure of confront are inactive of a person for reliable human face image reconstruction, the implementation system requires huge facial image datasets. Further the assessment of the system should be done by employing a testing procedure. This paper deals with the analysis and rectification of human face images for reconstruction and optimization of human face images. The advantage of using input human face image for reconstruction in forensic application and automatic face recognition system is that they are free from wide variety of poses, expression, illumination gestures and face occlusion. The whole research is divided into two phases; in the first phase reconstruction of destructed part of human face image is being done with template matching. Second phase deals with deep neural network applications to match the image carried out in phase one. The proposed algorithm is used to reconstruct the image and at the same time, reconstructed image is used as test image for biometrical face recognition. After reconstruction of image, it is examined with various well-known algorithms (SVMs, LDA, ICA, PCA & DNNs) of face recognition system for the evaluating the performance of speed, memory usage and metrics of accuracy.