Texture Classification on DTD (test)
91.7AccuracySFG (hat)
Evaluation results
| Method | Links | |
|---|---|---|
| SFG (hat)Kernel type=Sliced Figalli-Gigli (variant)2026.02 | 91.7 | |
| SFGKernel type=Sliced Figalli-Gigli2026.02 | 90.1 | |
| SWKernel type=Sliced Wasserstein2026.02 | 89.5 | |
| ZS_EnEvaluation Protocol=Zero-shot2023.11 | 49.35 | |
| TwistNet-18Params=11.59M, FLOPs=1.85G, Pre-training=None2026.02 | 45.8 | |
| Prototypical CurriculumShot=5-shot, Protocol=linear probing, Training epochs=4002024.11 | 45.59 | |
| CLIPBackbone=ViT-B/16, Evaluation Protocol=Zero-shot2023.11 | 44.56 | |
| BaselineShot=5-shot, Protocol=linear probing, Training epochs=16002024.11 | 44.47 | |
| CLIPBackbone=ViT-B/32, Evaluation Protocol=Zero-shot2023.11 | 43.79 | |
| Prototypical CurriculumShot=5-shot, Protocol=linear probing, Training epochs=2002024.11 | 42.81 | |
| FastViT-SA12Params=10.60M, FLOPs=1.50G, Pre-training=None2026.02 | 42.7 | |
| CLIPBackbone=RN50, Evaluation Protocol=Zero-shot2023.11 | 42.38 | |
| ResNet-18Params=11.20M, FLOPs=1.82G, Pre-training=None2026.02 | 39.4 | |
| RepViT-M1.5Params=13.67M, FLOPs=2.31G, Pre-training=None2026.02 | 39.2 | |
| CLIPBackbone=RN101, Evaluation Protocol=Zero-shot2023.11 | 38.48 | |
| SE-ResNet-18Params=11.29M, FLOPs=1.82G, Pre-training=None2026.02 | 36.7 | |
| Swin-TinyParams=27.56M, FLOPs=4.51G, Pre-training=None2026.02 | 32.2 | |
| ConvNeXtV2-NanoParams=15.01M, FLOPs=2.45G, Pre-training=None2026.02 | 29.1 | |
| ConvNeXt-TinyParams=27.86M, FLOPs=4.47G, Pre-training=None2026.02 | 11.1 |