Feature Point Detection and Description Networks Based on Asymmetric Convolution and the Cross-ResolutionImage-Matching Method

Table 3

HPatches homography estimation.

	Homography estimation
	= 3↑	= 5↑	= 10↑
SIFT	0.750	0.802	0.824
SURF	0.614	0.700	0.755
ORB	0.455	0.524	0.566
LIFT	0.689	0.698	0.722
LoFTR	0.659	0.756	0.846
Superpoint	0.753	0.821	0.845
DeepFEPE	0.752	0.822	0.851
ACPoint (ours)	0.778	0.845	0.871

The meanings in bold in Table 3 represent the values of the best performing methods for the indicators in this column. Meanwhile, the method proposed in this paper has the best performance in all indexes in this table.