FROZEN SCIENTIFIC BENCHMARK6 Models · 90 Runs
Six-Model Comparison
Side-by-side comparison of all six evaluated systems on 3 real EnMAP scenes × 5 seeds. Spatially separated splits, frozen test predictions.
Statistical interpretation
AgriSpectra-Q has the highest numerical mean F1 (96.40%). However, its advantage over HSI-RF is only +0.0008 with a 95% bootstrap CI of [−0.0012, +0.0027] — which crosses zero. Statistical superiority is not established.
METRIC
Mean F1 — all six models
TABLE 1
Six-Model Benchmark — Click header to sort
| Model | Mean F1 | PR-AUC | ROC-AUC | Brier ↓ | ECE ↓ |
|---|---|---|---|---|---|
AgriSpectra-QOur System | 96.40% | 99.47% | 99.87% | — | 0.73% |
Adaptive Classical | 96.34% | 99.47% | 99.86% | — | — |
HSI-RF | 96.31% | 99.49% | 99.87% | — | — |
48-band XGBoost | 95.22% | 99.23% | 99.80% | — | — |
Spectral XGBoost | 94.78% | 99.24% | 99.81% | — | — |
Current Hybrid | 89.98% | 96.16% | 98.75% | — | — |