A Partial-to-Partial Point Cloud Registration Method Based on Geometric Attention Network

Table 3

Results on 3DMatch and 3DLoMatch datasets with different sample sizes.

#Samples	3DMatch					3DLoMatch
	5,000	2,500	1,000	500	250	5,000	2,500	1,000	500	250
Registration recall (%)
3DSN [36]	78.4	76.2	71.4	67.6	50.8	33.0	29.0	23.3	17.0	11.0
FCGF [12]	85.1	84.7	83.3	81.6	71.4	40.1	41.7	38.2	35.4	26.8
D3Feat [13]	81.6	84.5	83.4	82.4	77.9	37.2	42.7	46.9	43.8	39.1
PREDATOR [15]	89.0	89.9	90.6	88.5	86.6	59.8	61.2	62.4	60.8	58.1
This study	89.0	88.8	89.3	88.7	86.0	56.1	57.5	57.8	56.3	53.5

Note. The best performance is highlighted in bold, while the next-best performance is underlined. FCGF, fully convolutional geometric features.