Table 3 Comparison of DACMixer with state-of-the-art methods on YMDA test set
From: Prediction of broken areas in murals based on MLP-fused long-range semantics
Model | Backbone | Broken. IoU[%] | MIoU[%] | Dice coefficient[%] |
|---|---|---|---|---|
GSCNN [26] | Wide-ResNet-38 [22] | 73.1 | 84.5 | 81.4 |
DANet [21] | ResNet-50 [32] | 61.3 | 77.6 | 76.0 |
HRNetv2 [27] | none | 66.7 | 80.8 | 80.0 |
OCRNet [28] | HRNetv2-W18 [27] | 67.0 | 81.1 | 80.3 |
GCNet [29] | ResNet-50 [32] | 62.4 | 78.1 | 76.8 |
Res-Unet [11] | ResNet-50 [32] | 66.8 | 80.7 | 76.8 |
TMCrack-Net [12] | ConvNext-S [34] | 71.2 | 81.8 | 78.2 |
STDC1 [30] | STDC1 [30] | 56.8 | 75.1 | 72.4 |
DPT [31] | Vit-Base [33] | 67.6 | 81.3 | 80.7 |
Ours | Wide-ResNet-38 [22] | 78.3 | 87.5 | 85.7 |