Face recognition has garnered significant attention over the last few decades due to its wide range of applications. It is used for access control to buildings and devices, tracking criminals, and many other purposes. However, building a face recognition system with high accuracy and low cost remains a challenge. With the advent of Convolutional Neural Networks (CNNs), which provide deep architectures for face recognition, developing such systems has become more accessible. Particularly, the use of transfer learning, where pre-trained models are leveraged for testing on specific datasets, has proven efficient in certain cases. This paper focuses on constructing a face recognition system based on the Viola-Jones detection algorithm combined with CNNs for recognition. The primary models evaluated in this work include AlexNet, VGG16, and ResNet. Among these, AlexNet achieved the best accuracy of 98% on the Georgia Tech dataset, outperforming other pre-trained models such as VGG16 and other state-of-the-art models.