The Feature Fusion Multi-Conv Networks using Best Feature Selection for Face Recognition
Abstract
This research paper presents a novel approach to face recognition through the implementation of Feature Fusion Multi-Conv Networks, utilizing the best feature selection techniques. The experiments were conducted using the FPFV, the MIT-CBCL, and the UFI dataset, where they were employed for pre-trained models of state-of-the-art CNNs, specifically ResNet-50 and SE-ResNet-50. The feature fusion process was implemented on the TensorFlow platform, utilizing Google Colab’s GPU support to enhance computational efficiency. The results indicate that the proposed method significantly improves face recognition accuracy compared to traditional techniques. Various recognizers, including KNN, MLP, LR, and SVM, were evaluated, showcasing high accuracy rates across different configurations. The results suggest that future research could focus on exploring additional feature extraction techniques and optimization methods to further enhance the robustness and performance of face recognition systems. This study contributes to the ongoing efforts in the field of computer vision and face recognition, aiming to develop faster, more accurate, and secure systems that can be practically implemented in real-world applications.