List your company
ResearchFormerly WizWand

Remote Sensing Scene Classification on RESISC45

98AccuracySatlasNet
18.12838.86459.680.336Nov 5, 2021Aug 6, 2022May 8, 2023Feb 7, 2024Nov 7, 2024Aug 9, 2025May 11, 2026
Updated 5mo ago

Evaluation results

MethodLinks
SatlasNetGeneral pre-training=-, Remote sensing pre-training=0.90M (SatlasPretrain), Parameters=88M (Swin-B), Evaluation Protocol=Fine-tuned2026.0598
SatMAE++General pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.70M (fMoW-Sentinel), Parameters=307M (ViT-L), Evaluation Protocol=Fine-tuned2026.0597.5
Multiple modelsBackbone=ViT-L/142022.0896.9
Multiple modelsBackbone=ViT-B/162022.0896.4
Joint patchingBackbone=ViT-L/142022.0896.3
DOFAGeneral pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.50M (Diverse), Parameters=307M (ViT-L), Evaluation Protocol=Fine-tuned2026.0596.1
Domain-adaptive pre-training2021.1195.89
Scale-MAE*General pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.40M (FMoW-RGB), Parameters=307M (ViT-L), Evaluation Protocol=Fine-tuned2026.0595.7
Foundation ModelGeneral pre-training=-, Remote sensing pre-training=0.28M, Parameters=30M (Swin-T), Evaluation Protocol=Fine-tuned2026.0595.59
Multiple modelsBackbone=ViT-B/322022.0895
ConvMAE*General pre-training=1.3M (ImageNet-1k), Remote sensing pre-training=-, Parameters=195M (ConvVit-L), Evaluation Protocol=Fine-tuned2026.0595
SatMAE*General pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.70M (fMoW-Sentinel), Parameters=307M (ViT-L), Evaluation Protocol=Fine-tuned2026.0594.8
PCNet2021.1194.59
GLDBS2021.1194.46
Seq. patchingBackbone=ViT-L/14, Seed=12022.0894.3
Million-AIDPre-training dataset=Million-AID2021.1194.26
Joint patchingBackbone=ViT-B/162022.0894.2
DOFAGeneral pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.50M (Diverse), Parameters=86M (ViT-B), Evaluation Protocol=Fine-tuned2026.0593.8
Seq. patchingBackbone=ViT-L/14, Seed=22022.0893.7
ResNet50+EANBackbone=ResNet502021.1193.51
MAEGeneral pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=-, Parameters=307M (ViT-L), Evaluation Protocol=Fine-tuned2026.0593.3
Joint patchingBackbone=ViT-B/322022.0892.6
DOFAGeneral pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.50M (Diverse), Parameters=307M (ViT-L), Evaluation Protocol=Frozen2026.0591.9
DOFAGeneral pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.50M (Diverse), Parameters=86M (ViT-B), Evaluation Protocol=Frozen2026.0591.3
Foundation ModelGeneral pre-training=-, Remote sensing pre-training=0.28M, Parameters=30M (Swin-T), Evaluation Protocol=Frozen2026.0591.27
Seq. patchingBackbone=ViT-L/14, Seed=02022.0890.8
Scale-MAE*General pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.40M (FMoW-RGB), Parameters=307M (ViT-L), Evaluation Protocol=Frozen2026.0589.6
Seq. patchingBackbone=ViT-B/16, Seed=12022.0889
MAEGeneral pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=-, Parameters=307M (ViT-L), Evaluation Protocol=Frozen2026.0588.9
SatMAE*General pre-training=14.0M (ImageNet-21k), Remote sensing pre-training=0.70M (fMoW-Sentinel), Parameters=307M (ViT-L), Evaluation Protocol=Frozen2026.0588.3
Seq. patchingBackbone=ViT-B/16, Seed=22022.0887.4
Seq. patchingBackbone=ViT-B/32, Seed=12022.0885.9
Parallel patchingBackbone=ViT-L/142022.0885.8
SatMAEViT=Large, Evaluation Protocol=Linear probing2025.0585
Seq. patchingBackbone=ViT-B/32, Seed=22022.0884.7
SatMAE++ViT=Large, Evaluation Protocol=Linear probing2025.0584.13
Seq. patchingBackbone=ViT-B/16, Seed=02022.0883.9
ConvMAE*General pre-training=1.3M (ImageNet-1k), Remote sensing pre-training=-, Parameters=195M (ConvVit-L), Evaluation Protocol=Frozen2026.0581.2
Seq. patchingBackbone=ViT-B/32, Seed=02022.0877.1
Parallel patchingBackbone=ViT-B/162022.0874.4
Sat2SoundViT=Base, Evaluation Protocol=Linear probing2025.0574.4
Parallel patchingBackbone=ViT-B/322022.0871.7
UnpatchedBackbone=ViT-L/142022.0871
UnpatchedBackbone=ViT-B/162022.0865.8
UnpatchedBackbone=ViT-B/322022.0860.7
xCLIPBackbone=ViT-B/16, Pre-trained on=IT35M, Mode=Zero-shot2022.1043.6
CLIPBackbone=ViT-B/16, Pre-trained on=IT35M, Mode=Zero-shot2022.1039.7
nCLIPBackbone=ViT-B/16, Pre-trained on=IT35M, Mode=Zero-shot2022.1021.2