Enhancing Face Detection: A Dual Model Approach with Modified VGG16 Layer

Face detection is a critical component in various intelligent applications, ranging from security systems to interactive user interfaces. Many face detection methods use deep learning and transfer learning models due to their superior accuracy and robustness. In this paper, we propose an approach based on VGG16 deep learning model tailored for dual output tasks: classification and regression. To enhance model performance, we first built an extensive dataset of facial images and applied data augmentation techniques to improve diversity and robustness. Then, we fine-tuned the VGG16 architecture to integrate dual output layers, optimizing the model for both tasks simultaneously. The proposed model is evaluated and compared to the standard VGG16 model under our dataset and another publicly available dataset. The modified VGG16 outperforms the standard model in terms of precision, accuracy, and execution time.