Skip to main content

Table 3 Comparisons of average precision (AP) on ResNet-50 and Faster R-CNN with FPN

From: Bridge-over-water detection via modulated deformable convolution and attention mechanisms

Backbones

\(\mathrm AP\)(%)

\(\mathrm AP^{s}\)(%)

\(\mathrm AP^{m}\)(%)

\(\mathrm AP^{l}\)(%)

Plain

68.4

48.4

72.1

85.8

Plain+Mdconv

69.3

48.6

73.9

87.2

Plain+Mdconv+CBAM

70.1

50.8

73.5

88.6

Plain+Mdconv+WCSA(CBAM)

69.7

51.2

73.6

86.4

Plain+Mdconv+SCSA(CBAM)

70.0

51.0

73.1

88.6

Pain+Mdconv+BAM

69.6

50.6

73.1

87.5

Plain+Mdconv+WCSA(BAM)

69.6

49.7

73.4

87.2

Plain+Mdconv+SCSA(BAM)

69.8

50.3

73.3

87.7

  1. The significance for bold values is only highlighted on the best performance