Semantic Segmentation on Cityscapes (~1/30 Subset, val)
76.2mIoUSemiVL
Evaluation results
| Method | Links | |
|---|---|---|
| SemiVLBackbone=ViT-B/162023.11 | 76.2 | |
| UniMatchBackbone=ViT-B/16, Note=re-produced2023.11 | 73.8 | |
| UniMatchBackbone=ResNet-1012023.11 | 73 | |
| ST++Architecture=DeepLabv3+, Backbone=ResNet-502021.06 | 61.4 | |
| PseudoSegArchitecture=DeepLabv3+, Backbone=ResNet-1012021.06 | 61 | |
| PseudoSegBackbone=ResNet-1012023.11 | 61 | |
| STArchitecture=DeepLabv3+, Backbone=ResNet-502021.06 | 60.9 | |
| CutMix-SegArchitecture=DeepLabv3+, Backbone=ResNet-1012021.06 | 55.7 | |
| SupOnlyArchitecture=DeepLabv3+, Backbone=ResNet-502021.06 | 55.1 | |
| DMTArchitecture=DeepLabv3+, Backbone=ResNet-1012021.06 | 54.8 | |
| CutMixBackbone=DeepLab v2, Pre-training=ImageNet, Labeled samples=1/30 (100), Number of runs=52019.06 | 51.2 | |
| CutoutBackbone=DeepLab v2, Pre-training=ImageNet, Labeled samples=1/30 (100), Number of runs=52019.06 | 47.21 | |
| Baseline (Ours)Backbone=DeepLab v2, Pre-training=ImageNet, Labeled samples=1/30 (100), Number of runs=52019.06 | 44.41 |