Semantic Segmentation on SYNTHIA to {Cityscapes, BDD, Mapillary, GTA} (test)
42.6mIoU (Cityscapes)TLDR
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TLDREncoder=ResNet-1012023.03 | 42.6 | 35.46 | 37.46 | 37.77 | — | |
| TLDREncoder=ResNet-502023.03 | 41.88 | 34.35 | 36.79 | 35.9 | — | |
| SAN-SAWEncoder=ResNet-1012023.03 | 40.87 | 35.98 | 37.26 | 30.79 | — | |
| FSDREncoder=ResNet-1012023.03 | 40.8 | 39.6 | 37.4 | — | — | |
| GTREncoder=ResNet-1012023.03 | 39.7 | 35.3 | 36.4 | 28.71 | — | |
| SAN-SAWEncoder=ResNet-502023.03 | 38.92 | 35.24 | 34.52 | 29.16 | — | |
| DRPCEncoder=ResNet-1012023.03 | 37.58 | 34.34 | 34.12 | 29.24 | — | |
| DRPCEncoder=ResNet-502023.03 | 35.65 | 31.53 | 32.74 | 28.75 | — | |
| ISWBackbone=ResNet-50, Source Dataset=Synthia (S), Training Iterations=40K, re-implementation=true2023.11 | 29.92 | 22.66 | 25.44 | 28.3 | 26.58 | |
| BaselineBackbone=ResNet-50, Source Dataset=Synthia (S), Training Iterations=40K2023.11 | 27.95 | 20.83 | 24.06 | 29.81 | 25.66 | |
| MRFPBackbone=ResNet-50, Source Dataset=Synthia (S), Training Iterations=40K2023.11 | 27.89 | 25.68 | 25.98 | 30.5 | 27.51 | |
| WildNetBackbone=ResNet-50, Source Dataset=Synthia (S), Training Iterations=40K, external_dataset_replaced_with_source=true2023.11 | 27.54 | 20.76 | 24.65 | 29.73 | 25.67 | |
| IBN-NetBackbone=ResNet-50, Source Dataset=Synthia (S), Training Iterations=40K, re-implementation=true2023.11 | 27.09 | 21.12 | 22.94 | 27.38 | 24.63 |