Face Recognition using SIFT and LBPH-based Hybrid Learning Model

The face recognition has been focused on object detection from images using various deep learning and machine learning models. This work proposes a deep learning model using multiple testing platforms to identify different face images from either videos or images. Two feature extraction methods, Scale-Invariant Feature Transform (SIFT) and Local Binary Patterns Histogram (LBPH) and a deep learning model (ResNet) is considered in the proposed model. LBPH features and ResNet models are considered for face recognition through videos, whereas the SIFT model is used for images. Since both deep learning and machine learning models have been considered for face recognition, their performance varied as per model when we used different videos and images. We have demonstrated our proposed model using real-time videos and images using the above technical approaches. LBPH efficiently detected and categorize local characteristics with an accuracy of 99.9%, whereas SIFT consistently maintained an accuracy rate of 88.9%. As our comparative observation, we found that the ResNet model performed well compared to other models.