Table 2 Segmentation accuracy of different semantic segmentation networks for Cabbage Seedling roots
From: Swin-Unet++: a study on phenotypic parameter analysis of cabbage seedling roots
Framework | Backbone | Acc↑(%) | mIoU↑(%) | Kappa↑(%) | Dice↑(%) | Params↓ (M) | Flops↓ (G) | mDT↓ (ms) |
|---|---|---|---|---|---|---|---|---|
Upernet | ResNet-101 | 97.62 | 84.12 | 87.65 | 90.60 | 93 | 513 | 27.2 |
DeeplabV3+ | Xception-65 | 97.49 | 83.47 | 87.02 | 90.13 | 46 | 191 | 25.3 |
Ocrnet | ResNet-101 | 97.48 | 83.31 | 83.89 | 90.02 | 45 | 162 | 62.4 |
FCN | HRNet-48 | 97.59 | 84.29 | 87.64 | 90.71 | 66 | 94 | 61.6 |
PSPnet | ResNet-101 | 97.62 | 83.75 | 87.54 | 90.34 | 87 | 343 | 26.0 |
Segnet | VGG-16 | 97.73 | 84.62 | 88.18 | 90.99 | 30 | 170 | 9.0 |
CCnet | ResNet-101 | 97.63 | 84.08 | 87.63 | 90.56 | 67 | 278 | 27.0 |
Unet | VGG-16 | 97.91 | 85.61 | 89.13 | 91.69 | 14 | 124 | 7.3 |
Unet++ | DenseNet | 97.93 | 85.77 | 89.26 | 91.79 | 8.3 | 119 | 13.0 |
Unet+++ | DenseNet | 97.82 | 85.10 | 88.63 | 91.29 | 27 | 792 | 10.7 |
Segformer | Transformer | 97.61 | 82.95 | 87.02 | 89.69 | 85 | 996 | 58.9 |
SwinUnet++ | Swintransformer | 98.19 | 86.69 | 90.37 | 92.38 | 60 | 354 | 29.5 |