Semisupervised Vector Quantization in Visual SLAM Using HGCN

Table 3

Comparing HGCN-FABMAP and FABMAP2 methods for LCD.

Train dataset	Test dataset	HGCN-FABMAP		FABMAP2
		Acc (%)	Rec (%)	Acc (%)	Rec (%)
St. Lucia	New College	50.07	100	50.02	100
311417	160500	56.00	86.00	56.8	82.00
		77.30	68.62	77.60	63.00
TUM sequence-11	Lip6indoor	50.30	100	50.34	100
392449	27189	54.90	90.50	54.90	90.50
		59.55	81.50	52.59	81.20
St. Lucia	St. Lucia	75.00	72.00	50.80	100
311417	235510	54.05	95.85	50.40	90.00
				59.55	81.50
New College	Newer College	90.20	80.85	71.00	80.30
160500	194000	97.40	72.34	79.00	72.34
		99.00	67.75	80.02	67.55