The increasing demand for advanced security systems incorporating facial recognition technology has led to the creation of various face identification techniques. However, only a select few have been proven to achieve optimal accuracy under specific situations. This study delved into an analysis of the main classifiers, which encompassed approaches rooted in swarm intelligence, support vector machines (SVMs), and deep learning. Specifically, the research introduces the binary Salp swarm algorithm (BSSA) and the binary grey wolf optimizer (BGWO) as novel pairwise classifiers for face identification. These two approaches, known as DT-BSSA and DT-BGWO, are organized in a 3-node binary decision tree, effectively reducing computing time. Additionally, the research explores the combination of CNN with a statistical classifier based on the salp swarm algorithm (SSA). Furthermore, various techniques for reducing feature space are examined, including PCA, LDA, and KFA. Comparative evaluation with decision tree-based fuzzy SVM classifier and modern methods of recognition using deep learning, such as CNN, shows that swarm-based classifiers displayed the highest efficacy in categorizing modestly sized databases. Nevertheless, the Deep learning solution emerged as the best choice in terms of speed and adaptability for applications requiring real-time processing.