Application Research for Fusion Model of Pseudolabel and Cross Network

Table 4

Experimental results of regular hyperparameters values.

(, , , )	VGG16 of CN [11]		ResNet50 of CN [12]		Single-task
	Facial Beauty dataset (%)	FaceShape dataset (%)	Facial Beauty dataset (%)	FaceShape dataset (%)	Facial Beauty Prediction (%)
(0, 1, 1, 0)	—	56.65	—	57.40	58.75
(0.1, 0.9, 0.9, 0.1)	62.35	58.91	62.95	56.35
(0.2, 0.8, 0.8, 0.2)	62.25	57.65	61.70	57.20
(0.3, 0.7, 0.7, 0.3)	62.25	56.85	62.60	57.45
(0.4, 0.6, 0.6, 0.4)	62.10	56.80	62.40	57.40
(0.5, 0.5, 0.5, 0.5)	61.80	57.05	61.55	56.80
(0.6, 0.4, 0.4, 0.6)	62.30	57.35	62.50	57.25
(0.7, 0.3, 0.3, 0.7)	62.40	57.60	62.05	57.90
(0.8, 0.2, 0.2, 0.8)	62.30	57.45	62.90	58.71
(0.9, 0.1, 0.1, 0.9)	62.80	57.10	62.70	58.20
(1, 0, 0, 1)	61.19	57.20	—	57.95

Better results in terms of recognition rates were highlighted in bold.