Incremental Learning On Imagenet 10 Steps
评估指标
# M Params
Average Incremental Accuracy
Average Incremental Accuracy Top-5
Final Accuracy
Final Accuracy Top-5
评测结果
各个模型在此基准测试上的表现结果
模型名称 | # M Params | Average Incremental Accuracy | Average Incremental Accuracy Top-5 | Final Accuracy | Final Accuracy Top-5 | Paper Title | Repository |
|---|---|---|---|---|---|---|---|
| DER w/o Pruning | 116.89 | 68.84 | 88.17 | 60.16 | 82.86 | DER: Dynamically Expandable Representation for Class Incremental Learning | |
| DyTox | 11.36 | 71.29 | 88.59 | 63.34 | 84.49 | DyTox: Transformers for Continual Learning with DYnamic TOken eXpansion | |
| E2E | 11.68 | - | 72.09 | - | 52.29 | End-to-End Incremental Learning | |
| WA | 11.68 | 65.67 | 86.60 | 55.60 | 81.10 | Maintaining Discrimination and Fairness in Class Incremental Learning | - |
| DER | - | 66.73 | 87.08 | 58.62 | 81.89 | DER: Dynamically Expandable Representation for Class Incremental Learning | |
| RMM (ResNet-18) | - | 67.45 | - | - | - | RMM: Reinforced Memory Management for Class-Incremental Learning | |
| BiC | 11.68 | - | 84.00 | - | 73.20 | Large Scale Incremental Learning | |
| FOSTER | - | 68.34 | - | - | - | FOSTER: Feature Boosting and Compression for Class-Incremental Learning | |
| kNN-CLIP | - | 85.5 | - | - | - | Revisiting a kNN-based Image Classification System with High-capacity Storage | - |
| iCaRL | 11.68 | 38.40 | 63.70 | 22.70 | 44.00 | iCaRL: Incremental Classifier and Representation Learning |
0 of 10 row(s) selected.