An Improved Face Detection Method Based on Face Recognition Application

Face recognition technology has been widely studied and applied for decades, and deep neural networks have greatly improved the indicators of face recognition systems. However, the face recognition application in reality is still subject to interference caused by direction, occlusion, shading and dynamic background, which makes the face recognition system unstable. This paper proposes an improved diagonal detection method, using a K parallel bottleneck connection structure, spacing parameters in each bottleneck connection structure, and using parameter partition sharing to reduce overfitting. The new loss function refines the difference between the detected corner point and the ground truth under different conditions, which can further improve the detection accuracy. Both the standard data set and the experiments in the real environment show that the proposed method has better detection accuracy and robustness.