Human Face Recognition for Attendance System Using Deep Learning Implementation with Convolutional Neural Network Method

Abstract

A Convolutional Neural Network-based deep learning approach is utilized to identify and classify various human facial images based on their characteristic differences. The facial recognition system utilizes a convolutional neural network (CNN) model implemented with TensorFlow, complemented by data augmentation techniques to prevent overfitting. The model is trained and validated using a specially prepared dataset, with performance evaluated through loss and accuracy graphs and confusion matrices. The system also supports real-time facial recognition using OpenCV, which identifies faces in videos and automatically records the subject's presence if the identification successfully passes a specified confidence threshold. This demonstrates the potential for integrating Deep Learning into automated facial recognition and attendance recording applications. Once the data training process is complete, the next stage is model testing to assess its performance. Data for testing is randomly sampled using the Python library. The results of this process will indicate the accuracy level of the method used. Confusion Matrix is used to calculate the accuracy value in Deep Learning. The testing process will be carried out through a classification process using the Convolutional Neural Network method. The accuracy of the model based on the Confusion Matrix is approximately 97%.