Multi-task Sequence Performance on CityScapes
-51.5Delta TbSPG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SPGBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | -51.5 | 5.5 | |
| SGPBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | -7.16 | 7.5 | |
| SPGBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | -5.8 | 5 | |
| SGPBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | -3.06 | 6 | |
| Vanilla trainingBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 0 | 6.5 | |
| Vanilla trainingBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 0 | 5.5 | |
| EWCBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 0.85 | 6.5 | |
| LwFBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 3.5 | 4.5 | |
| DERBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 3.99 | 5.5 | |
| iCaRLBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 4.76 | 5.5 | |
| HADBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 5.89 | 3.5 | |
| iCaRLBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 13.9 | 7 | |
| Joint trainingBackbone=ResNet-18, Task Sequence=Segmentation -> Depth2026.03 | 15.06 | 1 | |
| DERBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 18.92 | 3.5 | |
| EWCBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 19.98 | 5.5 | |
| LwFBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 24.91 | 5 | |
| HADBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 26.7 | 3.5 | |
| Joint trainingBackbone=ResNet-18, Task Sequence=Depth -> Segmentation2026.03 | 28.23 | 1 |