Age and Gender Estimation Using Shifting and Re-scaling of Local Regions - Technologies and Applications of Artificial Intelligence - page 10

Majority Voting

[1.0]

95.00%

[0.8, 1.0,
1.2]

92.88%

92.78%

91.05%

90.00%

85.00%

Fig. 7. The accuracy of gender classification from the voting of the different regions using
multiple scales

5.3 Age Estimation Results

For age estimation, we use the same algorithm as the gender classification algorithm
with using SVR instead of SVM. We measure the age estimation accuracy in mean
absolute error in years (lower is better).

Table 4. The accuracy of the age estimation on different regions separately using different
scales and a shift of 90 pixels

[1.0] [0.8, 1.0, 1.2]

Left Eyebrow

8.654 8.760

Right Eyebrow

8.931 8.873

Left Eye

7.804 7.493

Right Eye

7.745 7.613

Nose

7.766 7.491

Mouth

8.315 8.298

Face

7.955 7.785

Average 8.167 8.044