Fig 1.
Proposed system for real-time security surveillance.
Fig 2.
Using a simple image as the input.
Fig 3.
Extracted face from human image.
Fig 4.
Process of face detection.
Fig 5.
Architecture of RetinaFace.
Fig 6.
RetinaFace multi-task loss function.
Fig 7.
Triple loss function.
Fig 8.
Triple loss and selection.
Fig 9.
Sample image dataset for training.
Fig 10.
Single face detection with various poses.
Table 1.
Performance comparison between face detection techniques.
Fig 11.
Single face detection on different occlusion rate.
Table 2.
Performance comparison between different occlusion rates.
Fig 12.
Accuracy of facial recognition techniques on different λ value.
Fig 13.
Single face recognition result with various poses.
Fig 14.
Single face recognition on different occlusion rate.
Fig 15.
Recognize multiple human faces.
Table 3.
Performance comparison between face recognition techniques.
Table 4.
Performance comparison between baseline models of deep learning and the proposed method.