Table 7 Results of the MTL models comparison.

From: Real-time facial recognition via multitask learning on raspberry Pi

Model

Person accuracy (%)

Age accuracy (%)

Ethnicity accuracy (%)

MT-SENET-B21

88.9

91.1

94.1

37

91.6

96.1

97.5

MT-Vgg1638

94.8

96.5

98.7

MT-Xception38

96.5

96.1

97.3

39

92.1

97.3

97.5

MTL-InceptionV3

93.3

95.6

97.5

MTL-MobileNet

99.0

99.3

99.5

MTL-MobileNetV2

98.3

97.3

99

Back to article page