Performance Analysis of Face Verification through the Decision Fusion using Several Face Recognition Models

Face recognition is a critical aspect of security systems and personal identity. A facial recognition system should accurately detect the authenticity of a person’s face based on the given identity. This study introduces a decision fusion approach to enhance facial recognition accuracy by focusing on models implemented using the FaceScrub dataset. Various advanced convolutional neural network models, including the pre-trained VGG-Face, FaceNet, and OpenFace models, are utilized, with ResNet-50 being a custom-built addition to the Deepface library. The ResNet-50 model was trained on the FaceScrub dataset. The measurement standards include cosine and Euclidean distances. The experimental results highlight the impact of decision fusion. For example, the combination of the VGG-Face and FaceNet models using cosine distance, demonstrated an improved accuracy of 91.60% compared to the individual models. The findings of this study are expected to provide valuable insights into the effectiveness of decision fusion in enhancing facial recognition systems, with potential applications in security and identity verification.