Abstract
Face recognition is a key area in computer vision and artificial intelligence. With the advent of deep learning, novel methods have been developed that achieve high accuracy in this field. This study aims to enhance facial recognition accuracy by comparing two state-of-the-art algorithms – DeepFace and FaceNet – and proposing an ensemble approach that integrates their strengths. To this end, we conduct a comparative analysis of DeepFace’s deep convolutional neural network–based identification and FaceNet’s multidimensional vector embedding, then combine their output probabilities into a single ensemble model. Experimental evaluation on publicly available datasets under varying lighting conditions and head poses reveals that the ensemble consistently outperforms each individual algorithm, demonstrating superior accuracy and robustness to external factors. We conclude that this hybrid ensembling strategy significantly improves recognition performance, validating its potential for complex face matching tasks. These findings indicate that real-world applications – such as surveillance, identity verification, and biometric authentication – can benefit from adopting ensemble methods to achieve higher accuracy and resilience.