Table 6 Class-wise performance of best models in each method.
Task | Method | Metrics | Class | |||
|---|---|---|---|---|---|---|
Usable | Potential | Non-usable | Average | |||
Whole building façade | Swin Transformer | F1 score | 91.08 | 90.47 | 89.12 | 90.22 |
Accuracy | 92.85 | 91.92 | 90.57 | 91.78 | ||
ViT | F1 score | 89.82 | 89.03 | 87.93 | 88.93 | |
Accuracy | 91.67 | 90.76 | 89.32 | 90.58 | ||
PVT | F1 score | 89.27 | 88.52 | 87.25 | 88.35 | |
Accuracy | 91.02 | 90.22 | 88.73 | 89.99 | ||
MobileViT | F1 score | 88.48 | 87.73 | 86.07 | 87.43 | |
Accuracy | 90.22 | 89.27 | 87.88 | 89.12 | ||
Axial Transformer | F1 score | 87.98 | 87.22 | 85.68 | 86.96 | |
Accuracy | 89.78 | 88.88 | 87.42 | 88.69 | ||
First story | Swin Transformer | F1 score | 90.67 | 90.02 | 88.53 | 89.74 |
Accuracy | 93.87 | 92.48 | 90.52 | 92.29 | ||
ViT | F1 score | 89.27 | 88.53 | 87.08 | 88.29 | |
Accuracy | 92.53 | 91.12 | 89.12 | 90.92 | ||
PVT | F1 score | 88.72 | 87.88 | 86.47 | 87.69 | |
Accuracy | 92.03 | 90.72 | 88.92 | 90.56 | ||
MobileViT | F1 score | 87.83 | 86.97 | 85.43 | 86.74 | |
Accuracy | 91.23 | 89.82 | 87.87 | 89.64 | ||
Axial Transformer | F1 score | 87.18 | 86.32 | 84.87 | 86.12 | |
Accuracy | 90.83 | 89.57 | 87.37 | 89.26 | ||
