Long-tailed Classification on CUB (test)
60.1Accuracy PCDRAGON + Bal'Loss
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DRAGON + Bal'LossBackbone=ResNet-101, Method category=Late Fusion, Balanced loss=true2020.04 | 60.1 | 66.5 | |
| DRAGONBackbone=ResNet-101, Method category=Late Fusion2020.04 | 57.8 | 67.7 | |
| LAGOBackbone=ResNet-101, Method category=Multi-modal, Semantic-expert (SE)=true2020.04 | 54.8 | 66 | |
| MixtureBackbone=ResNet-101, Method category=Late Fusion2020.04 | 54.8 | 66 | |
| CB LWSBackbone=ResNet-101, Method category=Vision-only2020.04 | 53.1 | 65.7 | |
| CE LossBackbone=ResNet-101, Method category=Vision-only, Visual-expert (VE)=true2020.04 | 53 | 65.5 | |
| LDAM LossBackbone=ResNet-101, Method category=Vision-only2020.04 | 50.1 | 64.1 | |
| Range LossBackbone=ResNet-101, Method category=Vision-only2020.04 | 48.5 | 65.3 | |
| Anchor LossBackbone=ResNet-101, Method category=Vision-only2020.04 | 48.3 | 64.7 | |
| CADA-VAEBackbone=ResNet-101, Method category=Multi-modal2020.04 | 48.3 | 57.4 | |
| Focal LossBackbone=ResNet-101, Method category=Vision-only2020.04 | 46.4 | 61.3 |