Continual Semantic Segmentation on Pascal-VOC 15-1 (6 tasks) 2012
65.12mIoU (0-15)PLOP
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PLOPArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped2020.11 | 65.12 | 0.2111 | 0.5464 | 67.21 | |
| MiBArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped, Source=excerpted from [8]2020.11 | 35.1 | 0.135 | 0.297 | — | |
| MiBArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped, Source=re-implementation2020.11 | 34.22 | 0.135 | 0.2929 | 54.19 | |
| ILTArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped, Source=re-implementation2020.11 | 8.75 | 0.0799 | 0.0856 | 40.16 | |
| LwF-MCArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped, Source=re-implementation2020.11 | 6.4 | 0.084 | 0.069 | — | |
| ILTArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped, Source=excerpted from [8]2020.11 | 4.9 | 0.078 | 0.057 | — | |
| Joint2023.05 | 0.798 | 0.724 | 0.774 | — | |
| FairCLBackbone=SegFormer2023.05 | 0.735 | 0.228 | 0.615 | — | |
| FairCLBackbone=DeepLab-V32023.05 | 0.72 | 0.227 | 0.603 | — | |
| RCIL2023.05 | 0.706 | 0.237 | 0.594 | — | |
| PLOP2023.05 | 0.651 | 0.211 | 0.546 | — | |
| SDR2023.05 | 0.473 | 0.147 | 0.395 | — | |
| MiB2023.05 | 0.38 | 0.135 | 0.322 | — | |
| EWCArchitecture=DeepLab-V3, Backbone=ResNet-101, Pre-training=ImageNet, CSS Protocol=Overlapped, Source=re-implementation2020.11 | 0.3 | 0.043 | 0.013 | — | |
| ILT2023.05 | 0.096 | 0.078 | 0.092 | — | |
| LWF2023.05 | 0.06 | 0.039 | 0.055 | — |