Multipose Face Recognition-Based Combined Adaptive Deep Learning Vector Quantization

Table 4

The confusion matrix for the proposed CLVQ, ADLVQ, and CADLVQ based on SDUMLA-HMT and CASIA-V5 datasets in the training and testing phases.

DB = SDUMLA-HMT
	Training (318)			Testing (106)
	CLVQ	ADLVQ	CADLVQ	CLVQ	ADLVQ	CADLVQ
Sensitivity	93.44	95.16	97.72	95.00	96.19	99.02
Specificity	92.31	87.50	90.91	83.33	100.00	100.00
Precision	99.65	99.66	99.67	98.96	100.00	100.00
Accuracy	93.40	94.97	97.48	94.34	96.23	99.06
F1score	96.45	97.36	98.68	96.94	98.06	99.51
DB = CASIA-V5
	Training (1500)			Testing (1000)
Sensitivity	87.59	95.37	96.08	94.85	97.44	97.98
Specificity	92.31	93.75	90.91	66.67	80.00	100.00
Precision	99.17	99.86	99.86	98.92	99.48	100.00
Accuracy	88.00	95.33	96.00	94.00	97.00	98.00
F1score	93.02	97.56	97.93	96.84	98.45	98.98