<![CDATA[Face recognition has become a crucial topic in image processing and computer vision, particularly in university environments. This study explores the use of GhostFaceNet and YOLOv8 models to address the challenges of face recognition with a limited dataset, consisting of only one formal photo per individual. By applying image augmentation techniques, we improved the system's accuracy to 85%. GhostFaceNet excels in generating precise and detailed face embeddings, which are essential for accurate recognition. Meanwhile, YOLOv8 demonstrates superior speed in detecting faces under various lighting conditions and angles. Comparative results reveal that YOLOv8 achieves an accuracy of 81%, outperforming SSD, which only reaches 76%. Despite challenges related to the low quality of original images, the findings highlight the significant potential of deep learning-based face recognition systems. This research aims to compare SSD and YOLOv8 detection models using GhostFaceNet and contribute to the development of more effective and reliable face recognition methods in academic settings.]]>
Copyrights © 2024
