Face Keypoint Detection Method Based on Blaze_ghost Network

The accuracy and speed of facial keypoint detection are crucial factors for effectively extracting fatigue features, such as eye blinking yawning. This paper focuses on the improvement optimization algorithms, presenting a method based Blaze_ghost network providing more reliable support analysis. Firstly, is designed backbone with deeper structure parameters to better capture detail improving localization. Secondly, HuberWingloss loss function further reduce training difficulty model enhance its generalization ability. Compared traditional functions, can interference outliers (such noise occlusion) in training, improve model’s robustness complex situations, detection. Experimental results show that proposed achieves significant improvements both NME (Normal Mean Error) FR (Failure Rate) evaluation metrics. methods, demonstrates considerable localization while still maintaining high efficiency.