Face Recognition using MTCNN, Inception - Resnet with Ensemble Approach

One of the critical task included in the computer vision in face recognition. Security system, surveillance system, computer-human interactions are some of the applications where face recognition has been widely used. This paper proposes an novel face recognition model that inculcates deep learning and feature extraction techniques. Our model, named "Inception-Resnet" is built to efficiently and accurately recognize faces from images. Inception-Resnet employs a deep convolutional neural network and it also employs a deep CNN architecture that includes multiple layers of convolutional and pooling operations for robust feature extraction. The network is trained using a large-scale dataset consisting of labelled face images, that allows it to learn differential facial features. Additionally, Inception-resnet utilizes a loss function during training phase to encourage the network to learn compact and separable face embeddings in the feature space. Face verification is conducted using an ensemble approach which combines multiple classifiers, including Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and logistic regression, for the classification of face embeddings.