Semantic Segmentation on CityScapes (mIoU)
86.9mIoUState-of-the-Art (Liu et al., 2021)
Evaluation Results
| Method | Links | |
|---|---|---|
| State-of-the-Art (Liu et al., 2021)Mode=Absolute state-of-the-art2023.04 | 86.9 | |
| DINOv2Arch.=ViT-g/14, Evaluation=Multiscale (+ms)2023.04 | 81 | |
| DINOv2Arch.=ViT-L/14, Evaluation=Multiscale (+ms)2023.04 | 80.9 | |
| DINOv2Arch.=ViT-B/14, Evaluation=Multiscale (+ms)2023.04 | 80 | |
| DINOv2Arch.=ViT-S/14, Evaluation=Multiscale (+ms)2023.04 | 77.1 | |
| Ours-XL-tCategory=Ours, Backbone=SDXL2024.10 | 75.89 | |
| iBOTArch.=ViT-L/16, Evaluation=Multiscale (+ms)2023.04 | 74.5 | |
| Ours-XLCategory=Ours, Backbone=SDXL2024.10 | 74.47 | |
| Meta PromptsCategory=SOTA2024.10 | 71.94 | |
| Legacy-XLCategory=Baseline, Backbone=SDXL2024.10 | 71.67 | |
| DINOv2Arch.=ViT-g/14, Evaluation=Linear2023.04 | 71.3 | |
| FP BASELINEArchitecture=DeepLabv32024.05 | 70.7 | |
| OpenCLIPArch.=ViT-G/14, Evaluation=Multiscale (+ms)2023.04 | 70.3 | |
| DINOv2Arch.=ViT-L/14, Evaluation=Linear2023.04 | 70.3 | |
| DINOv2Arch.=ViT-B/14, Evaluation=Linear2023.04 | 69.4 | |
| BOLDArchitecture=DeepLabv3, Backbone=Boolean ResNet-182024.05 | 67.4 | |
| DINOv2Arch.=ViT-S/14, Evaluation=Linear2023.04 | 66.6 | |
| DINOArch.=ViT-B/8, Evaluation=Multiscale (+ms)2023.04 | 66.2 | |
| iBOTArch.=ViT-L/16, Evaluation=Linear2023.04 | 64.8 | |
| Ours-v1.5Category=Ours, Backbone=Stable Diffusion v1.52024.10 | 64.1 | |
| Legacy-v1.5Category=Baseline, Backbone=Stable Diffusion v1.52024.10 | 64.01 | |
| MAEArch.=ViT-H/14, Evaluation=Multiscale (+ms)2023.04 | 61 | |
| OpenCLIPArch.=ViT-G/14, Evaluation=Linear2023.04 | 60.3 | |
| MAEArch.=ViT-H/14, Evaluation=Linear2023.04 | 58.4 | |
| BINARY DAD-NET2024.05 | 58.1 | |
| DINOArch.=ViT-B/8, Evaluation=Linear2023.04 | 56.9 | |
| VPDCategory=SOTA2024.10 | 55.06 | |
| OursBackbone=SDv1.4, Resolution=512 x 5122026.06 | 27.8 | |
| DiffSegBackbone=SDv1.4, Resolution=512 x 5122026.06 | 22.2 | |
| EagleAX (Auxiliary image requirement)=true2026.06 | 22.1 | |
| STEGOLD (Language Dependency)=true, AX (Auxiliary image requirement)=true2026.06 | 21 | |
| ReCOLD (Language Dependency)=true, AX (Auxiliary image requirement)=true2026.06 | 19.3 | |
| MaskCut2026.06 | 18.7 | |
| DiffCutBackbone=SDv1.4, Resolution=512 x 5122026.06 | 13.4 | |
| PiCIEUA (Unsupervised Adaptation)=true2026.06 | 12.3 | |
| ProMerge2026.06 | 11.8 | |
| MaskCLIPLD (Language Dependency)=true2026.06 | 10 | |
| DINOv3 + Recursive-NCut2026.06 | 9 | |
| MDCUA (Unsupervised Adaptation)=true2026.06 | 7.1 | |
| IICUA (Unsupervised Adaptation)=true2026.06 | 6.4 | |
| CuVLERUA (Unsupervised Adaptation)=true2026.06 | 3.2 |