Classification 벤치마크
Classification on ImageNet C-OOD (class-out-of-distribution)
Detection AUROC (severity 0)
- 2023-02-23 — ViT-L/32-384 with Max-logit: Detection AUROC (severity 0) 0.9958
| Rank | Model | Detection AUROC (severity 0) | Detection AUROC (severity 5) | Detection AUROC (severity 10) | Paper | Code | Year |
|---|---|---|---|---|---|---|---|
| 1 | ViT-L/32-384 with Max-logit | 0.9958 | 0.9632 | 0.7748 | A framework for benchmarking class-out-of-distribution detection and its application to ImageNet | mdabbah/COOD_benchmarking | 2023 |
| 2 | ViT-L/32-384 with ODIN | 0.9955 | 0.9589 | 0.7635 | A framework for benchmarking class-out-of-distribution detection and its application to ImageNet | mdabbah/COOD_benchmarking | 2023 |
| 3 | ViT-L/32-384 with Entropy | 0.9948 | 0.9514 | 0.7332 | A framework for benchmarking class-out-of-distribution detection and its application to ImageNet | mdabbah/COOD_benchmarking | 2023 |
| 4 | ViT-L/32-384 with MC Dropout | 0.9947 | 0.9478 | 0.712 | A framework for benchmarking class-out-of-distribution detection and its application to ImageNet | mdabbah/COOD_benchmarking | 2023 |
| 5 | ViT-L/32-384 with Softmax | 0.9915 | 0.9293 | 0.7 | A framework for benchmarking class-out-of-distribution detection and its application to ImageNet | mdabbah/COOD_benchmarking | 2023 |