Exploring 2D and 3D Face Recognition with FaceNet across Multiple Classifiers

Face recognition technology has witnessed significant advancements in recent years, with researchers exploring novel approaches to enhance accuracy and robustness. This study investigates the performance of FaceNet, a state-of-the-art deep learning model for face recognition, in both 2D and 3D modalities. Additionally, various classifiers are employed to evaluate their impact on the overall recognition accuracy. Both publicly available as well as self-made datasets have been used in this study. The FaceNet model is employed to extract high-dimensional feature vectors from the facial data, leveraging its ability to map faces into a compact Euclidean space. The 2D face recognition system operates on conventional images, while the 3D counterpart processes depth information for a more comprehensive analysis. To evaluate the recognition accuracy, several classifiers are employed, including Support Vector Machines (SVM) and Logistic Regression (LR) model. The classifiers are trained and tested on both 2D and 3D datasets separately. Although, the effectiveness of FaceNet on 2D face recognition has been studied before, this article shows its effectiveness for 3D face recognition as well. Furthermore, the choice of classifier significantly influences the overall recognition accuracy. SVM exhibits robust performance across both modalities, showcasing its versatility in handling complex feature spaces. On the other hand, LR provides a better accuracy in cases where the dataset is equitably distributed.