Fig 1.
IIAAF overall framework (Image source: Yawning detection dataset (DOI: https://dx.doi.org/10.21227/e1qm-hb90), licensed under CC BY 4.0.).
Fig 2.
Neural network architecture for driver fatigue detection.
Table 1.
Experimental parameter settings.
Fig 3.
Sample images from Experiment 2 dataset (keyframes at different moments, image source: Yawning detection dataset (DOI: https://dx.doi.org/10.21227/e1qm-hb90), licensed under CC BY 4.0.).
Fig 4.
Training results of Algorithm 1 (Yawn dataset: Model detection results on the Yawn dataset; YawDDR dataset: Model detection results on the YawDDR dataset).
Fig 5.
Sample images from Experiment 2 dataset (left column indicates the degree of brightness reduction, right column indicates the degree of brightness increase, Image source: Yawning Detection Dataset (DOI: https://dx.doi.org/10.21227/e1qm-hb90), licensed under CC BY 4.0.)).
Fig 6.
Detection results of Experiment 2 (Yawn_B40%: Background brightness reduced by 40%, facial brightness increased by 40%. Yawn_B60%: Background brightness reduced by 60%, facial brightness increased by 60%. Yawn_B80%: Background brightness reduced by 80%, facial brightness increased by 80%).
Fig 7.
Sample images from Experiment 3 dataset (left column indicates the degree of brightness reduction, right column indicates the degree of brightness increase, Image source: Yawning Detection Dataset (DOI: https://dx.doi.org/10.21227/e1qm-hb90), licensed under CC BY 4.0.)).
Fig 8.
Detection results of Experiment 3 (YawnDDR_L20%: Background brightness reduced by 60%, facial brightness adjusted to 1.2x. YawnDDR_L40%: Background brightness reduced by 60%, facial brightness adjusted to 1.4x. YawnDDR_L60%: Background brightness reduced by 60%, facial brightness adjusted to 1.6x. YawnDDR_L80%: Background brightness reduced by 60%, facial brightness adjusted to 1.8x.).
Table 2.
Comparison of method’s accuracy.