TABLE 3.

Quantitative comparison results between the proposed method and other methods on the BSDS500 test set.

Method	ODS	OIS	AP	P(M)	FLOPs (G)
Human (Martin et al., 2004)	0.803	0.803	–	–	–
Canny (Canny, 1986)	0.611	0.676	0.520	–	–
SCO (Yang et al., 2015a)	0.670	0.710	0.710	–	–
SED (Akbarinia and Parraga, 2018)	0.710	0.740	0.740	–	–
gPb (Arbelaez et al., 2010)	0.729	0.755	0.745	–	–
OEF (Hallman and Fowlkes, 2015)	0.746	0.770	0.815	–	–
SE (Dollár and Zitnick, 2014)	0.743	0.764	0.800	–	–
DeepContour (Shen et al., 2015)	0.757	0.776	0.790	–	–
DeepEdge (Bertasius et al., 2015)	0.753	0.772	0.787	–	–
COB (Maninis et al., 2016)	0.793	0.819	0.849	28.8†	–
HED (Xie and Tu, 2015)	0.788	0.808	0.840	14.7‡	93.2‡
RCF-SS-VOC (Liu et al., 2017)	0.806	0.823	0.839	14.8‡	79.7‡
RCF-MS-VOC (Liu et al., 2017)	0.811	0.830	0.846
CED-SS (Wang et al., 2017)	0.803	0.820	0.871	21.4†	138.8†
CED-MS-VOC (Wang et al., 2017)	0.815	0.833	0.889
LPCB-SS-VOC (Deng et al., 2018)	0.808	0.824	–	15.7†	121.5†
LPCB-MS-VOC (Deng et al., 2018)	0.815	0.834	0.827
DRC-SS-VOC (Cao et al., 2020)	0.802	0.818	0.800	17.7‡	124.2‡
DRC-MS-VOC (Cao et al., 2020)	0.817	0.832	0.836
LRC-SS-VOC (Lin et al., 2020)	0.802	0.821	0.830	24.8‡	174.4‡
LRC-MS-VOC (Lin et al., 2020)	0.816	0.836	0.864
DSCD-SS-VOC (Deng and Liu, 2020)	0.813	0.836	0.847	34.07†	135.3†
DSCD-MS-VOC (Deng and Liu, 2020)	0.822	0.859	0.863
BDCN-SS-VOC (He et al., 2019)	0.820	0.838	0.888	16.3‡	95.1‡
BDCN-MS-VOC (He et al., 2019)	0.828	0.844	0.890
PiDiNet-SS-VOC (Su et al., 2021)	0.807	0.823	–	0.71‡	16.6‡
EDTER-SS-VOC (Pu et al., 2022)	0.832	0.847	0.886	–	332.0†
EDTER-MS-VOC (Pu et al., 2022)	0.848	0.865	0.903	–
MEDNet-SS-VOC	0.811	0.831	0.849	55.2‡	179.5‡
MEDNet-MS-VOC	0.825	0.845	0.872

VOC, mixed data set BSDS500-VOC; SS, single-scale results; MS, multi-scale results. The best two results are marked with red and blue, respectively. ^‡Stands for the result of our test. ^†Indicates the result of the reference. FLOPs are calculated based on a 320 × 320 image.