Class-incremental learning on Split CIFAR-100 (10-task)
95.02CAAHiDe-Prompt
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HiDe-PromptPTM=iBOT-1K, Backbone=ViT-B/162023.10 | 95.02 | 93.48 | — | — | — | 1 | |
| HiDe-PromptPTM=iBOT-21K, Backbone=ViT-B/162023.10 | 94.56 | 93.02 | — | — | — | 1.33 | |
| HiDe-PromptPTM=DINO-1K, Backbone=ViT-B/162023.10 | 94.25 | 92.51 | — | — | — | 0.99 | |
| HiDe-PromptPTM=Sup-21K, Backbone=ViT-B/162023.10 | 94.03 | 92.61 | — | — | — | 3.16 | |
| HiDe-PromptPTM=MoCo-1K, Backbone=ViT-B/162023.10 | 93.7 | 91.57 | — | — | — | 1.19 | |
| VQ-PromptBackbone=ViT-Base, Pre-training=ImageNet-1K2024.10 | 92.84 | 88.73 | — | — | — | — | |
| EvoPromptBackbone=ViT-Base, Pre-training=ImageNet-1K, Publication=AAAI242024.10 | 92.26 | 87.97 | — | — | — | — | |
| S-Prompt++PTM=Sup-21K, Backbone=ViT-B/162023.10 | 92.25 | 88.81 | — | — | — | 3.87 | |
| CODA-PromptPTM=Sup-21K, Backbone=ViT-B/162023.10 | 91.57 | 86.94 | — | — | — | 4.04 | |
| DualPromptPTM=Sup-21K, Backbone=ViT-B/162023.10 | 90.64 | 86.6 | — | — | — | 4.45 | |
| Deep L2P++Backbone=ViT-Base, Pre-training=ImageNet-1K, Publication=CVPR222024.10 | 90.5 | 84.3 | — | — | — | — | |
| L2P++Backbone=ViT-Base, Pre-training=ImageNet-1K, Publication=CVPR222024.10 | 88.96 | 82.5 | — | — | — | — | |
| L2PPTM=Sup-21K, Backbone=ViT-B/162023.10 | 88.25 | 83.06 | — | — | — | 6.58 | |
| CODA-PromptPTM=iBOT-21K, Backbone=ViT-B/162023.10 | 87.02 | 80.83 | — | — | — | 7.5 | |
| CODA-PromptPTM=iBOT-1K, Backbone=ViT-B/162023.10 | 86.21 | 79.11 | — | — | — | 7.69 | |
| DualPromptPTM=iBOT-21K, Backbone=ViT-B/162023.10 | 86.16 | 78.76 | — | — | — | 9.84 | |
| S-Prompt++PTM=iBOT-21K, Backbone=ViT-B/162023.10 | 85.85 | 79.14 | — | — | — | 9.17 | |
| S-Prompt++PTM=iBOT-1K, Backbone=ViT-B/162023.10 | 85.66 | 77.53 | — | — | — | 8.07 | |
| DualPromptPTM=MoCo-1K, Backbone=ViT-B/162023.10 | 85.31 | 77.77 | — | — | — | 6.61 | |
| L2PPTM=iBOT-21K, Backbone=ViT-B/162023.10 | 85.13 | 79 | — | — | — | 5.55 | |
| DualPromptPTM=iBOT-1K, Backbone=ViT-B/162023.10 | 85.08 | 76.63 | — | — | — | 8.41 | |
| CODA-PromptPTM=MoCo-1K, Backbone=ViT-B/162023.10 | 84.97 | 76.83 | — | — | — | 12.6 | |
| CODA-PromptPTM=DINO-1K, Backbone=ViT-B/162023.10 | 84.81 | 77.5 | — | — | — | 8.1 | |
| DualPromptPTM=DINO-1K, Backbone=ViT-B/162023.10 | 83.98 | 74.9 | — | — | — | 10.26 | |
| S-Prompt++PTM=MoCo-1K, Backbone=ViT-B/162023.10 | 83.88 | 76.3 | — | — | — | 14.67 | |
| S-Prompt++PTM=DINO-1K, Backbone=ViT-B/162023.10 | 83.82 | 74.97 | — | — | — | 7.78 | |
| L2PPTM=MoCo-1K, Backbone=ViT-B/162023.10 | 83.14 | 74.85 | — | — | — | 6.51 | |
| L2PPTM=iBOT-1K, Backbone=ViT-B/162023.10 | 82.69 | 75.57 | — | — | — | 7.23 | |
| L2PPTM=DINO-1K, Backbone=ViT-B/162023.10 | 79.02 | 70.65 | — | — | — | 9.46 | |
| DualPromptBackbone=ViT-Base, Pre-training=ImageNet-1K, Publication=ECCV222024.10 | 77.92 | 66 | — | — | — | — | |
| FT++Backbone=ViT-Base, Pre-training=ImageNet-1K2024.10 | 74.76 | 49.91 | — | — | — | — | |
| CODA-PBackbone=ViT-Base, Pre-training=ImageNet-1K, Publication=CVPR232024.10 | 74.26 | 70.03 | — | — | — | — | |
| FTBackbone=ViT-Base, Pre-training=ImageNet-1K2024.10 | 37.37 | 29.21 | — | — | — | — | |
| Batch Model ConsolidationBuffer Size=20002023.05 | 6.27 | — | 66.5 | — | 78 | — | |
| ERBuffer Size=10002023.05 | 5.4 | — | 12.4 | — | 184 | — | |
| GDumbBuffer Size=50002023.05 | 4.29 | — | 36 | — | 129 | — | |
| EWCon2023.05 | 0.99 | — | 13.1 | — | 172 | — | |
| A-GEMBuffer Size=50002023.05 | 0.67 | — | 6.5 | — | 231 | — | |
| iCaRLBuffer Size=10002023.05 | 0.59 | — | 25.5 | — | 141 | — | |
| DER++Buffer Size=50002023.05 | 0.53 | — | 27 | — | 205 | — | |
| DMC2023.05 | 0.02 | — | 36.2 | — | 140 | — | |
| GSSBuffer Size=50002023.05 | — | — | 17.4 | — | 1,203 | — | |
| Joint-Train.Backbone=ViT-Base, Pre-training=ImageNet-1K2024.10 | — | 91.38 | — | — | — | — | |
| LwFBackbone=ViT-Base, Pre-training=ImageNet-1K, Publication=TPAMI172024.10 | — | 64.83 | — | — | — | — | |
| Multi-Task2023.05 | — | — | 75.79 | — | 100 | — | |
| SGD2023.05 | — | — | 8.5 | — | 100 | — |