Incremental Learning 벤치마크
Incremental Learning on CIFAR100B020Step(5ClassesPerStep)
Average Incremental Accuracy
- 2021-03-31 — DER(ResNet-18): Average Incremental Accuracy 73.98
- 2022-12-29 — TCIL-Lite: Average Incremental Accuracy 75.47
- 2025-03-24 — View-Batch(DER): Average Incremental Accuracy 76.95
| Rank | Model | Average Incremental Accuracy | Paper | Code | Year |
|---|---|---|---|---|---|
| 1 | View-Batch(DER) | 76.95 | Do Your Best and Get Enough Rest for Continual Learning | hankyul2/viewbatchmodel | 2025 |
| 2 | TCIL-Lite | 75.47 | Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning | yellowpancake/tcil | 2022 |
| 3 | TCIL | 75.11 | Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning | yellowpancake/tcil | 2022 |
| 4 | DER(ResNet-18) | 73.98 | DER: Dynamically Expandable Representation for Class Incremental Learning | g-u-n/pycil · Rhyssiyan/DER-ClassIL.pytorch | 2021 |
| 5 | FOSTER | 70.65 | FOSTER: Feature Boosting and Compression for Class-Incremental Learning | g-u-n/pycil · G-U-N/ECCV22-FOSTER | 2022 |