Table 1 Regression loss weights on the AFLW2000 and BIWI datasets.

From: Soft-label guided stacked dual attention network for head pose estimation and its application to classroom gaze analysis

\(\alpha\)

AFLW2000

BIWI

Yaw

Pitch

Roll

MAE

Yaw

Pitch

Roll

MAE

4

3.11

4.30

2.67

3.36

4.20

4.10

2.42

3.57

2

2.94

4.36

2.69

3.33

3.83

3.98

2.32

3.38

1

2.96

4.50

2.64

3.36

3.69

4.13

2.36

3.40

0.1

3.00

4.39

2.66

3.35

3.63

4.25

2.45

3.44

Back to article page