Published Paper Details:

Facial Recognition (FR) is a technique that helps recognize people's faces and has wide applications. New methodologies based on computer vision, deep learning, and transfer learning have been used to perform face recognition. However, limitations still exist. This study proposes a comprehensive methodology that evaluates multiple Convolutional Neural Network (CNN) and transfer learning-based models, including VGG16, InceptionV3, and CBAM-enhanced CNNs, us ing a Kaggle celebrity face dataset. The VGG16-based model achieved the highest validation accuracy (88.65%) and lowest loss (0.4267), while a custom CNN (256 units, 128x128) offered a balance of accuracy (85.13%) and compu tational efficiency. Challenges such as overfitting and computational cost were observed, particularly in regularized models. Future work aims to enhance generalization through data augmentation, hyperparameter tuning, and model ensemble techniques to address scalability and real-world distortions.