Semantic Segmentation on Drought (Poyang Lake, China) Whole-scene
95.1Pixel Accuracy (PA)ViT+UNet, 2-stage
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ViT+UNet, 2-stageArchitecture=ViT+UNet, Number of stages=2-stage, Label supervision=No truth label2025.07 | 95.1 | 88.4 | 84.5 | |
| PCA + KMeans (k=2)Label supervision=Expert-refined labels, K value=22025.07 | 94.7 | 81.2 | 74.9 | |
| EVAPLabel supervision=No truth label2025.07 | 91.4 | 84.7 | 81.5 | |
| SVMLabel supervision=Expert-refined labels2025.07 | 85.4 | 67.6 | 59.7 | |
| PCA + SVMLabel supervision=Expert-refined labels2025.07 | 85.4 | 67.6 | 59.7 |