Interpretation on CUB (val)
3.4ADGrad-CAM
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Grad-CAMDNN architecture=Swin Transformer V2-B2024.04 | 3.4 | 50.1 | 29.2 | 56.3 | 96.79 | 4 | |
| CAMDNN architecture=Swin Transformer V2-B2024.04 | 3.5 | 49.7 | 27.7 | 54.9 | 96.79 | 3 | |
| Layer-CAMDNN architecture=Swin Transformer V2-B2024.04 | 3.5 | 48.4 | 28.6 | 56.4 | 97.06 | 2 | |
| μ-CAPE (PF)DNN architecture=Swin Transformer V2-B, mode=Probabilistic Ensemble (PF)2024.04 | 4.1 | 52.7 | 40.3 | 55 | 94.25 | 6 | |
| Grad-CAM++DNN architecture=Swin Transformer V2-B2024.04 | 4.2 | 47.5 | 26.3 | 58.4 | 95.69 | 0 | |
| μ-CAPE (TS)DNN architecture=Swin Transformer V2-B, mode=Teacher Softening (TS)2024.04 | 4.2 | 47.6 | 37.6 | 54 | 96.96 | 1 | |
| Score-CAMDNN architecture=Swin Transformer V2-B2024.04 | 6.5 | 46.1 | 46.5 | 78.1 | 43.95 | 7 | |
| CAPE (PF)DNN architecture=Swin Transformer V2-B, mode=Probabilistic Ensemble (PF)2024.04 | 14.3 | 33.7 | 22.7 | 71.4 | 17.19 | 4 | |
| CAPE (TS)DNN architecture=Swin Transformer V2-B, mode=Teacher Softening (TS)2024.04 | 21.9 | 22.6 | 19.6 | 73.4 | 21.34 | 4 |