Table 3 Comparison of PRE@0.03 accross all test samples with MobileNetV323, InceptionV324, ResNet25, SENet26, DRN27, ConvNeXt28, TransNeXt29 and our method as backbones in three scenarios.
From: Multi-scale feature fusion keypoint detection network for ship draft line localization
Backbone | Scene | ||
|---|---|---|---|
PRE@0.03 \(\downarrow\) | |||
Daytime (%) | Nighttime (%) | Colored Ships (%) | |
MobileNetV3 | 11.67 | 9.38 | 12.50 |
InceptionV3 | 18.30 | 15.63 | 12.50 |
ResNet-34 | 36.67 | 31.25 | 37.50 |
SENet-34 | 51.67 | 53.13 | 62.50 |
DRN-34 | 73.33 | 62.50 | 75.00 |
ConvNeXt | 76.67 | 71.88 | 87.50 |
TransNeXt | 83.33 | 81.25 | 87.50 |
MFFKD-34 | 91.67 | 93.75 | 100.00 |
