Table 1 Key module ablation experiments.
From: Scene image visual layout based on deep encoder–decoder network and visual image attention model
Methods | ResNet-50 | HED | VGAT | EEFM | DRRM | IoU | FCR (%) | IT (ms) | FLOPs(G) |
|---|---|---|---|---|---|---|---|---|---|
DEDN-VGAT LayoutGen | ✓ | ✓ | ✓ | ✓ | ✓ | 0.82 | 94.64 | 26.94 | 49.10 |
DEDN-VGAT LayoutGen | × | ✓ | ✓ | ✓ | ✓ | 0.71 | 86.12 | 28.21 | 53.70 |
DEDN-VGAT LayoutGen | ✓ | × | ✓ | ✓ | ✓ | 0.75 | 89.34 | 24.53 | 47.90 |
DEDN-VGAT LayoutGen | ✓ | ✓ | × | ✓ | ✓ | 0.68 | 85.76 | 34.12 | 61.20 |
DEDN-VGAT LayoutGen | ✓ | ✓ | ✓ | × | ✓ | 0.76 | 90.05 | 29.87 | 50.80 |
DEDN-VGAT LayoutGen | ✓ | ✓ | ✓ | ✓ | × | 0.74 | 88.92 | 27.45 | 51.30 |
DEDN-VGAT LayoutGen | ✓ | ✓ | ✓ | × | × | 0.69 | 84.31 | 32.56 | 58.60 |
TS-SLG | – | – | – | – | – | 0.68 | 87.72 | 31.90 | 59.83 |
StSG | – | – | – | – | – | 0.59 | 83.61 | 47.59 | 71.63 |
SA-GAT | – | – | – | – | – | 0.73 | 80.30 | 29.20 | 73.87 |
AWSP-JO | – | – | – | – | – | 0.64 | 83.28 | 42.74 | 76.86 |
