Table 6 Performance of super-resolution models for image classification across multiple countries (super-resolved and original).
Algeria | Egypt | Ghana | Malawi | Uganda | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
SR models | Accuracy | F1 score | Kappa | Accuracy | F1 score | Kappa | Accuracy | F1 score | Kappa | Accuracy | F1 score | Kappa | Accuracy | F1 score | Kappa |
DBPISR | 69% | 51.40% | 44.45% | 71% | 64.56% | 56.28% | 48% | 38.25% | 34.1% | 57% | 42.64% | 32.2% | 53.34% | 34.5% | 31.16% |
BSRGAN | 85% | 72.64% | 79.87% | 51% | 50.22% | 58.4% | 4% | 5.77% | 11.76% | 60% | 58.17% | 58.84% | 5.33% | 6.89% | 0% |
Real-ESRGAN | 96% | 78.31% | 92.33% | 68% | 62.94% | 64.5% | 18.67% | 22.05% | 17.10% | 79% | 68.91% | 71% | 18.67% | 21.8% | 15.9% |
SwinIR | 94% | 77.09% | 88.04% | 64% | 60.42% | 64.09% | 12% | 15.61% | 7.45% | 84% | 72.52% | 75% | 10.67% | 13.69% | 5.6% |
SwinIR_large | 84% | 72.39% | 80% | 63% | 59.75% | 66.6% | 5.33% | 7.28% | 4.73% | 59% | 56.8% | 64.13% | 17.33% | 20.45% | 15.68% |
Original | 85% | 72.74% | 80.67% | 62% | 59.5% | 62.33% | 5.33% | 7.54% | 2.96% | 66% | 62.37% | 62.8% | 18.67% | 21.87% | 13.95% |
