Table 2 Validation of subjective and objective image layout quality.
From: Scene image visual layout based on deep encoder–decoder network and visual image attention model
Metrics | DEDN-VGAT LayoutGen | TS-SLG | SA-GAT | Professor’s |
|---|---|---|---|---|
Subjectivity | ||||
 Layout rationality score | 9.42 ± 0.21 | 8.10 ± 0.37 | 8.55 ± 0.32 | 9.10 ± 0.26 |
 Layout consistency score | 9.38 ± 0.23 | 8.15 ± 0.45 | 8.52 ± 0.36 | 9.05 ± 0.28 |
 Semantic consistency score | 9.43 ± 0.18 | 8.25 ± 0.42 | 8.63 ± 0.35 | 9.20 ± 0.24 |
Objectivity | ||||
 Bounding box overlap rate | 3.8% ± 0.7% | 9.2% ± 1.5% | 6.9% ± 1.3% | 3.5% ± 0.6% |
 Time-averaged IoU | 0.91 ± 0.02 | 0.79 ± 0.06 | 0.83 ± 0.05 | 0.89 ± 0.03 |
 Key feature retention rate | 95.2% ± 1.8% | 85.8% ± 3.6 | 89.1% ± 2.9 | 93.6% ± 2.2 |
