List your company
ResearchFormerly WizWand

Semantic Segmentation on SpectralEarth Benchmarks Aggregate

1.4RankSpectralEarth-FM-a
1.0163.6086.28.792May 20, 2026
Updated 5mo ago

Evaluation results

MethodLinks
SpectralEarth-FM-aSensor branch routing=multi-branch, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.051.4
SpectralEarth-FM-sSensor branch routing=single-branch, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.052.5
Spec. ViT-LModel scale=Large, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.052.8
Spec. ViT-BModel scale=Base, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.053.7
PanopticonEvaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.054.9
HyperSIGMA-BModel scale=Base, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.056.8
CARLEvaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.057.3
HyperSIGMA-LModel scale=Large, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.058.1
DOFA-BModel scale=Base, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.058.3
DOFA-LModel scale=Large, Evaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.059.2
SpecAwareEvaluation protocol=frozen encoder, Segmentation head type=SegFormer-style2026.0511