Table 6 Comparison of segmentation evaluation indexes between the Res-UNet-CA and other segmentation methods. Bold face indicates the best performance.
From: A high-precision segmentation method based on UNet for disc cutter holder of shield machine
Model | Accuracy | Precision | Recall | F1_Score | IoU | mIoU |
|---|---|---|---|---|---|---|
FCN | 0.9346 | 0.8771 | 0.8883 | 0.8827 | 0.7900 | 0.8516 |
LR-ASPP | 0.7315 | 0.5113 | 0.7050 | 0.5927 | 0.4212 | 0.5437 |
DeepLabV3 | 0.8997 | 0.8922 | 0.7256 | 0.8003 | 0.6672 | 0.7708 |
DeepLabV3+ | 0.9499 | 0.9871 | 0.8303 | 0.9019 | 0.8213 | 0.8782 |
PSP-DANet | 0.9642 | 0.9503 | 0.9188 | 0.9343 | 0.8767 | 0.9143 |
SegFormer | 0.9302 | 0.8758 | 0.8733 | 0.8746 | 0.7771 | 0.8424 |
Res-UNet-CA | 0.9945 | 0.9890 | 0.9911 | 0.990 | 0.9803 | 0.9863 |
