MRBENet: A Multiresolution Boundary Enhancement Network for Salient Object Detection

Table 2

Ablation experiment for feature fusion.

Method	SOD			HKU-IS			PASCAL-S
	MAE↓	MaxF↑	S↑	MAE↓	MaxF↑	S↑	MAE↓	MaxF↑	S↑
(a) ADD	0.0991	0.8672	0.7926	0.0307	0.9363	0.9186	0.0658	0.8717	0.8616
(b) Multiplication	0.1002	0.8679	0.7945	0.0309	0.9366	0.9188	0.0666	0.8739	0.8614
(c) Concat + conv	0.0997	0.8707	0.7943	0.0301	0.9365	0.9185	0.0663	0.8711	0.8605
(d) ADD + SA	0.0979	0.8670	0.7955	0.0307	0.9365	0.9186	0.0650	0.8761	0.8613
(e) Our model	0.0966	0.8695	0.8019	0.0299	0.9354	0.9181	0.0643	0.8764	0.8613