Semantic Segmentation on PRISMA Tiled dataset (test)
44.37F1Fine-tune (PRISMA)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Fine-tune (PRISMA)dataset=PRISMA, epochs=62025.11 | 44.37 | 37.05 | 40.32 | 50.14 | |
| Ensembled Zero-shotmode=Ensemble, training=Zero-shot2025.11 | 44.04 | 35.6 | 37.02 | 54.36 | |
| Ensembled Fine-tune (EMIT+PRISMA)mode=Ensemble, dataset=EMIT+PRISMA, epochs=22025.11 | 43.79 | 36.08 | 37.97 | 51.72 | |
| Ensembled Fine-tune (PRISMA)mode=Ensemble, dataset=PRISMA, epochs=62025.11 | 43.61 | 40.61 | 43.11 | 44.13 | |
| Fine-tune (EMIT+PRISMA)dataset=EMIT+PRISMA, epochs=22025.11 | 42.11 | 33 | 34.36 | 54.71 | |
| Zero-shot (EMIT U-Net RGB+WMF model)backbone=U-Net RGB+WMF, pre-training=EMIT2025.11 | 40.5 | 31.84 | 31.57 | 57.18 | |
| Train from scratchepochs=502025.11 | 31.55 | 25.91 | 31 | 33.61 | |
| Baseline (WMF 500 thr, “ones” kernel)threshold=500, kernel=ones2025.11 | 26.05 | — | 23.14 | 29.8 |