Face recognition system is based on combining conventional methods and the computerized system that uses deep learning methods to automate the process of attendance recording. The current method does not give precision in the real-life educational context as the brightness, direction, blockage and image quality vary. To overcomes this problem, a novel model proposes employing a Convolutional Neural Network (CNN) with spatial attention to identify essential facial characteristics and detect students, with an objective of boosting accuracy, minimizing time consumption and enhancing overall efficiency and reliability of attendance management in educational settings. A Wiener and Median (WieMed) Filter are utilized for preprocessing for improving the acquired image from the camera which becomes the input data for the CNN. The face is detected using You Only Look Once (YOLO) and it is extracted by CNN for efficient attendance marking of the student. The extracted features are of the face uses the PCA for feature reduction then Euclidian Distance Clustering recognize the face of the student the attendance is stored in the Excel sheet. The proposed model achieved an accuracy of 97.50% for recognizing the face. The proposed model improves overall accuracy by 25.07%, 7.82%, 1.15%, and 7.82% compared to the Eigen Face Recognizer algorithm, CNN, MTCNN, FaceNet, and Deep learning approach, correspondingly.