Face Recognition (FR), is one of the biometric techniques which is used to recognize any given face image. Several mechanisms have been recommended for FR; however, in real-time situations, it remains very tedious. In order to differentiate people, a primary approach relies on several conditions like posture variety, illumination, facial occlusion. To resolve these issues in the conventional methodologies, the Deep Learning (DL)-centric Efficient Face Recognition System (FRS) using Z-Normalization along with the Moore Penrose-centric Deep Convolutional Neural Network (ZMP-DCNN) algorithm is presented in this paper. The input image is enhanced by using the BIG-Weber technique to enhance the contrast. Cost Function-grounded You Only Look Once version 3 (CF-YOLOv3) algorithm is used to detect the face as of the enhanced image. Then, to localize as well as to represent the face’s salient regions, the facial points are extracted using the Multi-Gated Supervised Descent (MGSD) approach. Next, the facial parts are segmented as of the detected face using the Angle rotation Adaptive Viola-Jones Algorithm (A2VJA). The significant features are selected as of the extracted features using the Newtonian Constant of Gravitation-based Grasshopper Optimization Algorithm (NCGOA). To recognize the face, the selected features were given as input to the ZMP-DCNN framework. A similar process of input images was done for the query images. The query image is recognized using trained outcome of the input image. Finally, performance of the proposed technique is compared with the existing frameworks. Thus, in contrast to the other mechanisms, the proposed face recognition system achieves superior performance.