Table 2 Experimental results of comparing methods in term of the MAE, SM, \(F_\beta\) and \(E_\epsilon\), where the best result in each column is marked in boldface. Note: \(\uparrow\) and \(\downarrow\) indicate that larger is better and smaller is better, respectively.

From: Triple-attentions based salient object detector for strip steel surface defects

Methods	\(\rho =0\)				\(\rho =10\%\)				\(\rho =20\%\)
Methods	MAE\(\downarrow\)	\(F_\beta\)\(\uparrow\)	\(E_\epsilon\)\(\uparrow\)	SM\(\uparrow\)	MAE\(\downarrow\)	\(F_\beta\)\(\uparrow\)	\(E_\epsilon\)\(\uparrow\)	SM\(\uparrow\)	MAE\(\downarrow\)	\(F_\beta\)\(\uparrow\)	\(E_\epsilon\)\(\uparrow\)	SM\(\uparrow\)
2LSG²²	0.2474	0.3892	0.561	0.5516	0.2587	0.3694	0.543	0.5367	0.2619	0.3597	0.5371	0.5336
BASNet²⁶	0.0152	0.9083	0.9691	0.9284	0.016	0.9021	0.9675	0.924	0.016	0.9007	0.9678	0.9234
BC¹⁶	0.1554	0.4296	0.6455	0.5943	0.1519	0.4184	0.6231	0.5882	0.1753	0.3739	0.6103	0.5624
BMPM⁵²	0.0299	0.8482	0.901	0.8586	0.0496	0.6587	0.8003	0.7449	0.0524	0.6348	0.7727	0.7337
CPD²¹	0.0211	0.8746	0.9393	0.9039	0.0353	0.7828	0.8864	0.831	0.0324	0.7876	0.8925	0.8384
DACNet²	0.0118	0.925	0.9773	0.9428	0.0125	0.9223	0.9762	0.9378	0.0144	0.9111	0.9679	0.9246
DSS¹⁹	0.0239	0.8088	0.855	0.8244	0.0453	0.6292	0.7339	0.7218	0.0432	0.6028	0.6972	0.7131
EDRNet¹²	0.013	0.9204	0.9754	0.9381	0.0139	0.9098	0.9725	0.9304	0.0146	0.9043	0.9701	0.9253
\(\hbox {F}^3\)Net⁵³	0.015	0.9088	0.9719	0.9216	0.0265	0.8194	0.9344	0.852	0.0276	0.8064	0.9233	0.8454
ITSD⁵⁴	0.0153	0.8968	0.9691	0.9248	0.0244	0.8239	0.9428	0.8656	0.0262	0.8099	0.9293	0.859
MIL¹⁸	0.1824	0.4513	0.6017	0.6186	0.2128	0.374	0.546	0.5685	0.2083	0.3785	0.5461	0.577
MINet⁵⁵	0.0144	0.9049	0.967	0.9242	0.0276	0.8065	0.9313	0.8535	0.0234	0.8409	0.9382	0.8716
NLDF⁵⁶	0.0484	0.704	0.8075	0.803	0.1171	0.439	0.6542	0.6345	0.1244	0.3683	0.5904	0.5893
PFANet²³	0.0848	0.5924	0.7478	0.7412	0.1011	0.5246	0.716	0.6958	0.1075	0.485	0.6591	0.6778
PiCANet²⁴	0.0263	0.8325	0.9237	0.8968	0.0355	0.7683	0.8879	0.8499	0.0404	0.7256	0.8616	0.823
PoolNet²⁵	0.0215	0.8512	0.9347	0.9031	0.0345	0.7451	0.8752	0.8215	0.0372	0.7179	0.8581	0.8104
\(\hbox {R}^3\)Net²⁰	0.0255	0.8315	0.9071	0.8395	0.0382	0.7225	0.8342	0.7665	0.043	0.6461	0.7557	0.7271
RCRR⁵⁷	0.2439	0.343	0.5493	0.5343	0.2552	0.3264	0.5279	0.5303	0.2842	0.3081	0.5133	0.5146
SAMNet⁵⁸	0.0218	0.8631	0.94	0.9064	0.0229	0.8514	0.9357	0.8987	0.0235	0.845	0.9334	0.8948
SMD¹⁷	0.2045	0.4268	0.5934	0.5843	0.1994	0.4212	0.5904	0.5846	0.1981	0.3973	0.5775	0.5718
Ours	0.0091	0.94	0.9846	0.9506	0.0091	0.94	0.9846	0.9506	0.0123	0.9163	0.9763	0.9315

Back to article page

Table 2 Experimental results of comparing methods in term of the MAE, SM, \(F_\beta\) and \(E_\epsilon\), where the best result in each column is marked in boldface. Note: \(\uparrow\) and \(\downarrow\) indicate that larger is better and smaller is better, respectively.

Search

Quick links