Deep Learning-Based Celebrity Face Recognition Using Convolutional Neural Networks with Image Augmentation

Abstract

Celebrity face recognition has become an important research area in computer vision due to its wide range of applications in biometric authentication, security systems, and digital media analysis. Advances in deep learning, particularly Convolutional Neural Networks (CNNs), have enabled automated face recognition by learning highly discriminative visual features directly from image data. This study aims to develop a CNN-based celebrity face recognition model using a custom dataset consisting of 1,200 facial images categorized into four classes: Blackpink, Ariana Grande, BTS, and Taylor Swift. Following image validation and selection, the dataset was divided into training and testing subsets with an 80:20 ratio, resulting in 70 images for model evaluation. The preprocessing pipeline included image resizing to 128 × 128 pixels, pixel value normalization, contrast enhancement, and data augmentation techniques, including random rotations of up to ±20°, 20% translation, 20% zoom, horizontal flipping, and brightness variation. The CNN model was trained for 30 epochs using the Adam optimizer and categorical cross-entropy loss function. Experimental results showed that the proposed model achieved an overall classification accuracy of 63%. Among the four classes, the Blackpink class obtained the highest F1-score of 0.79, whereas the Taylor Swift class recorded the lowest F1-score of 0.29. These findings indicate that the CNN model is capable of learning fundamental facial characteristics for celebrity recognition; however, further improvements in model architecture, dataset diversity, and optimization strategies are required to enhance recognition performance and classification robustness.