Table 1 Comparisons of average precision (AP) on DarkNet-53 and YOLOv3
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 | 48.9 | 36.3 | 63.3 | 46.4 |
Plain+Mdconv | 58.7 | 37.7 | 64.1 | 76.0 |
Plain+Mdconv+CBAM | 62.1 | 43.4 | 66.1 | 78.9 |
Plain+Mdconv+WCSA(CBAM) | 61.6 | 44.2 | 64.7 | 78.2 |
Pain+Mdconv+SCSA(CBAM) | 62.8 | 43.3 | 66.6 | 80.9 |
Plain+Mdconv+BAM | 62.1 | 43.2 | 66.1 | 78.7 |
Plain+Mdconv+WCSA(BAM) | 62.1 | 42.7 | 66.0 | 80.2 |
Plain+Mdconv+SCSA(BAM) | 62.6 | 44.5 | 66.0 | 79.2 |