Scene Classification on NWPU-RESISC45 (test)
96.74Top-1 AccuracyMS-PoE + BAIR
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MS-PoE + BAIRBackbone=EarthDial2026.05 | 96.74 | — | — | — | 75.37 | 1.68 | |
| LWGANet-L2Params (M)=13, FLOPs (G)=1.87, Training Image Size=224x224, Batch Size=GPU: 256, CPU: 16, ARM: 322025.01 | 96.17 | 3,308 | 16.18 | 274.3 | — | — | |
| LWGANet-L1Params (M)=5.9, FLOPs (G)=0.709, Training Image Size=224x224, Batch Size=GPU: 256, CPU: 16, ARM: 322025.01 | 95.7 | 6,418 | 34.08 | 375.8 | — | — | |
| LWGANet-L0Params (M)=1.72, FLOPs (G)=0.186, Training Image Size=224x224, Batch Size=GPU: 256, CPU: 16, ARM: 322025.01 | 95.49 | 13,234 | 80 | 687.8 | — | — | |
| MobileNet V2 1.0xParams (M)=2.28, FLOPs (G)=0.319, Training Image Size=224x224, Batch Size=GPU: 256, CPU: 16, ARM: 322025.01 | 95.06 | 11,301 | 49.11 | 785.4 | — | — | |
| MobileViT XSParams (M)=2.02, FLOPs (G)=0.9, Training Image Size=256x256, Batch Size=GPU: 256, CPU: 16, ARM: 322025.01 | 94.9 | 3,300 | 12.99 | 306.3 | — | — | |
| MS-PoEBackbone=EarthDial2026.05 | 94.69 | — | — | — | 69.03 | 3.41 | |
| LongLLMLingua + BAIRBackbone=EarthDial2026.05 | 94.49 | — | — | — | 66.04 | 3.41 | |
| LongLLMLinguaBackbone=EarthDial2026.05 | 94.38 | — | — | — | 63.81 | 3.36 | |
| FasterNet T0Params (M)=2.68, FLOPs (G)=0.338, Training Image Size=224x224, Batch Size=GPU: 256, CPU: 16, ARM: 322025.01 | 93.3 | 18,276 | 106.4 | 839.7 | — | — | |
| BaselineBackbone=EarthDial2026.05 | 93.13 | — | — | — | — | — | |
| Visual-focus Instr. + BAIRBackbone=EarthDial2026.05 | 92.62 | — | — | — | 62.69 | 5.18 | |
| Visual-focus Instr.Backbone=EarthDial2026.05 | 92.51 | — | — | — | 63.81 | 5.37 | |
| Standard RAG + BAIRBackbone=EarthDial2026.05 | 91.72 | — | — | — | 59.7 | 5.92 | |
| MAD-RAGBackbone=EarthDial2026.05 | 91.64 | — | — | — | 62.69 | 6.22 | |
| MAD-RAG + BAIRBackbone=EarthDial2026.05 | 91.64 | — | — | — | 62.31 | 6.19 | |
| Standard RAGBackbone=EarthDial2026.05 | 91.28 | — | — | — | 60.82 | 6.47 | |
| LongLLMLingua + BAIRBackbone=SkySenseGPT-7B2026.05 | 70.79 | — | — | — | 46.41 | 18.2 | |
| LongLLMLinguaBackbone=SkySenseGPT-7B2026.05 | 69.97 | — | — | — | 46.5 | 19.43 | |
| BaselineBackbone=SkySenseGPT-7B2026.05 | 68.9 | — | — | — | — | — | |
| MAD-RAG + BAIRBackbone=SkySenseGPT-7B2026.05 | 67.13 | — | — | — | 46.17 | 23.41 | |
| Visual-focus Instr. + BAIRBackbone=SkySenseGPT-7B2026.05 | 66.74 | — | — | — | 46 | 23.89 | |
| Visual-focus Instr.Backbone=SkySenseGPT-7B2026.05 | 66.59 | — | — | — | 46.83 | 24.49 | |
| Standard RAG + BAIRBackbone=SkySenseGPT-7B2026.05 | 66.31 | — | — | — | 44.02 | 23.63 | |
| MAD-RAGBackbone=SkySenseGPT-7B2026.05 | 66.18 | — | — | — | 45.67 | 24.56 | |
| MS-PoE + BAIRBackbone=SkySenseGPT-7B2026.05 | 66.05 | — | — | — | 44.27 | 24.12 | |
| Standard RAGBackbone=SkySenseGPT-7B2026.05 | 65.74 | — | — | — | 44.68 | 24.75 | |
| MS-PoEBackbone=SkySenseGPT-7B2026.05 | 65.36 | — | — | — | 45.26 | 25.57 |