Incremental Learning on ImageNet - 10 steps
Metric: Average Incremental Accuracy (higher is better)
Metric: Average Incremental Accuracy (higher is better)
| # | Model↕ | Average Incremental Accuracy▼ | Augmentations | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | kNN-CLIP | 85.5 | Yes | Revisiting a kNN-based Image Classification Syst... | 2022-04-03 | - |
| 2 | DyTox | 71.29 | No | DyTox: Transformers for Continual Learning with ... | 2021-11-22 | Code |
| 3 | DER w/o Pruning | 68.84 | No | DER: Dynamically Expandable Representation for C... | 2021-03-31 | Code |
| 4 | FOSTER | 68.34 | No | FOSTER: Feature Boosting and Compression for Cla... | 2022-04-10 | Code |
| 5 | RMM (ResNet-18) | 67.45 | No | RMM: Reinforced Memory Management for Class-Incr... | 2023-01-14 | Code |
| 6 | DER | 66.73 | No | DER: Dynamically Expandable Representation for C... | 2021-03-31 | Code |
| 7 | WA | 65.67 | No | Maintaining Discrimination and Fairness in Class... | 2019-11-16 | Code |
| 8 | iCaRL | 38.4 | No | iCaRL: Incremental Classifier and Representation... | 2016-11-23 | Code |