Class Incremental Semantic Segmentation on LoveDA domain-incremental protocol
63.7mIoU (All)ProtoFlow
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ProtoFlowBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 63.7 | 85.6 | 3.1 | |
| Joint (oracle)Backbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 63 | 85.1 | 3.5 | |
| MiRBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 61.8 | 84 | 4.2 | |
| STCL-DRNetBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 61.1 | 83.4 | 4.6 | |
| CoGaMiDBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 60.9 | 83.7 | 4.5 | |
| MiSSNetBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 60.5 | 83.2 | 4.8 | |
| GSMF-RS-DILBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 60.5 | 83.6 | 4.7 | |
| APRBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 60 | 82.8 | 5 | |
| LAGBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 59.3 | 82.1 | 5.3 | |
| DKDBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 57.8 | 80.5 | 6 | |
| MiBBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 56.2 | 79.6 | 6.8 | |
| Fine-tuneBackbone=HRNet-W48, Pre-training=ImageNet-1K2026.04 | 49.5 | 72.9 | 11.9 |