Continual Semantic Segmentation on ADE20k 50-50 (3 tasks) (val)
38mIoU (51-150)Joint
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| JointArchitecture=Mask-Based2022.11 | 38 | — | 43.1 | — | 53.4 | |
| Joint2023.05 | 32.8 | 51.1 | 38.9 | — | — | |
| FairCLBackbone=SegFormer2023.05 | 27.78 | 49.62 | 35.15 | 42.25 | — | |
| MiB + AWT2023.05 | 26.85 | 46.6 | 33.5 | — | — | |
| FairCLBackbone=DeepLab-V32023.05 | 26.84 | 49.65 | 34.55 | 41.68 | — | |
| PLOPArchitecture=Mask-Based2022.11 | 26.6 | — | 33.8 | 36.5 | 48.1 | |
| CoMFormerArchitecture=Mask-Based2022.11 | 26.6 | — | 34.1 | 36.6 | 49.2 | |
| MiBArchitecture=Per-Pixel2022.11 | 25.4 | — | 31.9 | 35 | 44.9 | |
| RCILArchitecture=DeepLab-v32022.11 | 25 | — | 32.5 | — | 48.3 | |
| RCIL2023.05 | 25 | 48.3 | 32.5 | — | — | |
| PLOPArchitecture=Per-Pixel2022.11 | 24.7 | — | 30.9 | 34.6 | 43.2 | |
| MiBArchitecture=DeepLab-v32022.11 | 21.6 | — | 29.3 | 38.9 | 45.3 | |
| PLOPArchitecture=DeepLab-v32022.11 | 21.6 | — | 30.4 | 39.4 | 48.6 | |
| MiBBackbone=ResNet-101, Architecture=Deeplab-V3, Pre-training=ImageNet, CSS Setting=Overlapped2020.11 | 21.01 | 45.57 | 29.31 | 38.98 | — | |
| MiB2021.06 | 21.01 | 45.57 | 29.31 | 38.98 | — | |
| MiB2023.05 | 21.01 | 45.57 | 29.31 | 38.98 | — | |
| PLOPBackbone=ResNet-101, Architecture=Deeplab-V3, Pre-training=ImageNet, CSS Setting=Overlapped2020.11 | 20.99 | 48.83 | 30.4 | 39.42 | — | |
| PLOP2021.06 | 20.99 | 48.83 | 30.4 | 39.42 | — | |
| PLOP2023.05 | 20.99 | 48.83 | 30.4 | 39.42 | — | |
| SSUL2023.05 | 20.15 | 48.38 | 29.56 | — | — | |
| MiBArchitecture=Mask-Based2022.11 | 19.4 | — | 21.1 | 25.8 | 24.6 | |
| FTArchitecture=Mask-Based2022.11 | 13.3 | — | 8.9 | 12.8 | 0 | |
| ILTBackbone=ResNet-101, Architecture=Deeplab-V3, Pre-training=ImageNet, CSS Setting=Overlapped2020.11 | 12.85 | 3.53 | 9.7 | 30.12 | — | |
| ILT2021.06 | 12.85 | 3.53 | 9.7 | 30.12 | — | |
| ILT2023.05 | 12.85 | 3.53 | 9.7 | 30.12 | — |