Table 4 Comparison of different backbones on RFRB
From: Automatic rape flower cluster counting method based on low-cost labelling and UAV-RGB images
Methods | Acc | MAE | rMAE | rMSE | rrMSE | R\(^2\) | Model capacity | |
|---|---|---|---|---|---|---|---|---|
Backbone | Loss | |||||||
\(Mnasnet0\_75\) | Bayesian loss | 0.7100 | 53.99 | 7.35 | 65.08 | 47.90 | 0.8278 | 20.8MB |
Densenet121 | 0.8679 | 27.38 | 5.23 | 32.58 | 17.13 | 0.9568 | 53.2MB | |
\(Efficientnet\_b3\) | 0.9047 | 28.34 | 5.32 | 38.89 | 11.61 | 0.9385 | 58.7MB | |
Vgg19 | 0.9098 | 22.90 | 4.78 | 31.95 | 12.15 | 0.9623 | 86.0MB | |
RapeNet | 0.8981 | 25.31 | 5.03 | 32.73 | 13.21 | 0.9566 | 4.9MB | |
\(RapeNet+\) | 0.9062 | 23.65 | 4.86 | 29.95 | 12.03 | 0.9635 | 5.6MB | |