Face recognition is a critical aspect of computer vision applications, with numerous algorithms continually evolving to improve accuracy and efficiency. In this study, a comprehensive analysis of twelve face recognition algorithms on a diverse dataset has been conducted, employing preprocessing techniques such as face cropping and grayscale conversion for enhanced efficiency. This analysis included traditional methods like principal component analysis (PCA), linear discriminant analysis (LDA), local binary patterns (LBPs), histogram of oriented gradients (HOG), and various support vector machine (SVM) kernels, alongside state-of-the-art deep learning models including DeepID, VGGFace2, CASIA-WebFace, ResNet50, and ResNet42. Surprisingly, CASIA-WebFace exhibited superior performance with an accuracy of 98.89%, accompanied by a “Time to build model” of 43.28 ms and “Time to predict outcomes” of 0.169 ms. Recognizing the significance of computational efficiency, further optimization of the ResNet algorithm has been done, achieving a 41% reduction in “Time to build model” and a 36% decrease in “Time to predict outcomes” at the expense of a 2.73% accuracy drop.