Publication

Abstract : Real-time multi-face tracking and clustering are the crucial roles now in modern surveillance, security, and identity verification systems. This paper presents a hybrid approach that integrates Histogram of Oriented Gradients (HOG) for feature extraction, Convolutional Neural Networks (CNN) for deep learning-based feature representation, and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for dynamic face clustering. The system efficiently handles variations in lighting and occlusions, as well as large-scale datasets. To evaluate performance, we compare our approach with state-of-the-art deep learning models, including YOLO and SSD, for memory efficiency, accuracy, and computational requirements. Experimental results are presented with high accuracy, robust clustering performance, and computational efficiency, making it suitable for real-world applications. The research also compares different clustering techniques and evaluates performance using standard metrics such as precision-recall trade-off, false acceptance rate, and frame rate analysis. This method allows for optimal computational efficiency as well as the capability to run real-time accurately and thus could be applied for surveillance and biometric identity verification in dynamic environments.