A Symmetric Fusion Learning Model for Detecting Visual Relations and Scene Parsing

Table 1

Comparison with state-of-the-art on the VG data set.

Method	SGGen			SGCls			PredCls
	R@20	R@50	R@100	R@20	R@50	R@100	R@20	R@50	R@100
IMP [29]	14.6	20.7	24.5	31.7	34.6	35.4	52.7	59.3	61.3
Frequency [31]	17.7	23.5	27.6	27.7	32.4	34.0	49.4	59.9	64.1
Frequency + overlap [31]	20.1	26.2	30.1	29.3	32.3	32.9	53.6	60.6	62.2
MotifNet-LeftRight [31]	21.4	27.2	30.3	32.9	35.8	36.5	58.5	65.2	67.1
Graph R-CNN [39]	—	11.4	13.7	—	29.6	31.6	—	54.2	59.1
VCTREE-SL [40]	21.7	27.7	31.1	35.0	37.9	38.6	59.8	66.2	67.9
RelDN [37]	21.1	28.3	32.7	36.1	36.8	36.8	66.9	68.4	68.4
VCTREE + TranstextNet[41]	—	28.1	31.7	—	38.3	39.3	—	66.9	68.7
Ours