EADF-Net: FGGS Grassland Champion Model
Official weights and verified benchmark evaluation for EADF-Net on the Fine-Grained Grassland Segmentation (FGGS) benchmark.
🏆 Benchmark Results vs Published Springer Paper
Reference Paper: The Visual Computer, Springer Nature, 2026 (Table 6).
| Metric / Tier | SGDC-DeepLab (Paper Table 6) | EADF-Net Ensemble (Ours) | Difference |
|---|---|---|---|
| Overall mIoU | 51.48% | 52.53% | +1.05% 🏆 |
| Overall Accuracy (OA) | 71.40% | 72.18% | +0.78% 🏆 |
| Overall mF1 | 66.21% | 66.88% | +0.67% 🏆 |
| Middle Grassland | 32.12% | 33.91% | +1.79% 🏆 |
| Low Grassland | 68.84% | 69.27% | +0.43% 🏆 |
| High Grassland | 79.51% | 79.22% | -0.29% |
📦 Files
- : PyTorch model weights (88.8 MB).
- : Complete precision, recall, and IoU breakdown.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support