Segmentation on ADE20K
55.3mIoUEfficientSAM
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| EfficientSAMVariant=S2025.06 | 55.3 | — | — | — | — | |
| STTVariant=L2025.06 | 55.3 | — | — | — | — | |
| STTVariant=H2025.06 | 55.2 | — | — | — | — | |
| SAMVariant=B2025.06 | 54.7 | — | — | — | — | |
| EfficientSAMVariant=Ti2025.06 | 54.4 | — | — | — | — | |
| SAMVariant=H2025.06 | 54.3 | — | — | — | — | |
| STTVariant=B2025.06 | 54.1 | — | — | — | — | |
| MogaNet-XLFramework=UperNet (160K), Crop size=640x640, Param. (M)=214, FLOPs (G)=2451, IN-21K pre-trained=true2022.11 | 54 | — | — | — | — | |
| SAMVariant=L2025.06 | 53.7 | — | — | — | — | |
| ConvNeXt-LFramework=UperNet (160K), Crop size=640x640, Param. (M)=245, FLOPs (G)=2458, IN-21K pre-trained=true2022.11 | 53.7 | — | — | — | — | |
| InfoMamba2026.03 | 53 | — | — | — | — | |
| SegMAN-B2026.03 | 52.6 | — | — | — | — | |
| RepLKNet-31LFramework=UperNet (160K), Crop size=640x640, Param. (M)=207, FLOPs (G)=2404, IN-21K pre-trained=true2022.11 | 52.4 | — | — | — | — | |
| Swin-LFramework=UperNet (160K), Crop size=640x640, Param. (M)=234, FLOPs (G)=2468, IN-21K pre-trained=true2022.11 | 52.1 | — | — | — | — | |
| VWFormer-B52026.03 | 52 | — | — | — | — | |
| SegNeXt-L2026.03 | 51 | — | — | — | — | |
| SegFormer-B52026.03 | 51 | — | — | — | — | |
| MogaNet-LFramework=UperNet (160K), Crop size=512x512, Param. (M)=113, FLOPs (G)=11762022.11 | 50.9 | — | — | — | — | |
| SLaK-BFramework=UperNet (160K), Crop size=512x512, Param. (M)=135, FLOPs (G)=11852022.11 | 50.2 | — | — | — | — | |
| MogaNet-BFramework=UperNet (160K), Crop size=512x512, Param. (M)=74, FLOPs (G)=10502022.11 | 50.1 | — | — | — | — | |
| RepLKNet-31BFramework=UperNet (160K), Crop size=512x512, Param. (M)=112, FLOPs (G)=11702022.11 | 49.9 | — | — | — | — | |
| Swin-BFramework=UperNet (160K), Crop size=512x512, Param. (M)=121, FLOPs (G)=11882022.11 | 49.7 | — | — | — | — | |
| SLaK-SFramework=UperNet (160K), Crop size=512x512, Param. (M)=91, FLOPs (G)=10282022.11 | 49.4 | — | — | — | — | |
| MogaNet-SFramework=UperNet (160K), Crop size=512x512, Param. (M)=55, FLOPs (G)=9462022.11 | 49.2 | — | — | — | — | |
| ConvNeXt-BFramework=UperNet (160K), Crop size=512x512, Param. (M)=122, FLOPs (G)=11702022.11 | 49.1 | — | — | — | — | |
| EDAFormer-B2026.03 | 49 | — | — | — | — | |
| ConvNeXt-SFramework=UperNet (160K), Crop size=512x512, Param. (M)=82, FLOPs (G)=10272022.11 | 48.7 | — | — | — | — | |
| HorNet-T7x7Framework=UperNet (160K), Crop size=512x512, Param. (M)=52, FLOPs (G)=9262022.11 | 48.1 | — | — | — | — | |
| Swin-SFramework=UperNet (160K), Crop size=512x512, Param. (M)=81, FLOPs (G)=10382022.11 | 48.1 | — | — | — | — | |
| FeedFormer-B22026.03 | 48 | — | — | — | — | |
| MogaNet-SFramework=Semantic FPN (80K), Crop size=512x512, Param. (M)=29, FLOPs (G)=1892022.11 | 47.7 | — | — | — | — | |
| UniFormer-SFramework=UperNet (160K), Crop size=512x512, Param. (M)=52, FLOPs (G)=10082022.11 | 47.6 | — | — | — | — | |
| MobileSAM2025.06 | 47.1 | — | — | — | — | |
| VAN-B2Framework=Semantic FPN (80K), Crop size=512x512, Param. (M)=30, FLOPs (G)=1642022.11 | 46.7 | — | — | — | — | |
| ConvNeXt-TFramework=UperNet (160K), Crop size=512x512, Param. (M)=60, FLOPs (G)=9392022.11 | 46.7 | — | — | — | — | |
| Uniformer-SFramework=Semantic FPN (80K), Crop size=512x512, Param. (M)=25, FLOPs (G)=2472022.11 | 46.6 | — | — | — | — | |
| Swin-TFramework=UperNet (160K), Crop size=512x512, Param. (M)=60, FLOPs (G)=9452022.11 | 46.1 | — | — | — | — | |
| Semantic Twins-SFramework=Semantic FPN (80K), Crop size=512x512, Param. (M)=28, FLOPs (G)=1622022.11 | 44.3 | — | — | — | — | |
| LITV2-SFramework=Semantic FPN (80K), Crop size=512x512, Param. (M)=31, FLOPs (G)=1792022.11 | 44.3 | — | — | — | — | |
| OmniGenTraining Setting=Multi-task2024.12 | 44.23 | — | — | — | — | |
| DeiT-SFramework=UperNet (160K), Crop size=512x512, Param. (M)=52, FLOPs (G)=10992022.11 | 44 | — | — | — | — | |
| ControlNet++Training Setting=Single-task2024.12 | 43.64 | — | — | — | — | |
| UNIC-AdapterTraining Setting=Multi-task2024.12 | 42.89 | — | — | — | — | |
| Swin-TFramework=Semantic FPN (80K), Crop size=512x512, Param. (M)=32, FLOPs (G)=1822022.11 | 41.5 | — | — | — | — | |
| PVT-SFramework=Semantic FPN (80K), Crop size=512x512, Param. (M)=28, FLOPs (G)=1612022.11 | 39.8 | — | — | — | — | |
| BaselineArchitecture=MobileNetV2 with DeepLabV3, #ReLUs (K)=852622025.12 | 34.08 | — | — | — | — | |
| Spatial sharingArchitecture=MobileNetV2 with DeepLabV3, #ReLUs (K)=25582025.12 | 33.23 | — | — | — | — | |
| DeepShareArchitecture=MobileNetV2 with DeepLabV3, #ReLUs (K)=25562025.12 | 33.22 | — | — | — | — | |
| ControlNetTraining Setting=Single-task2024.12 | 32.55 | — | — | — | — | |
| OFTBase Model=Stable Diffusion v1.5, Training Epochs=202023.06 | 27.06 | — | — | 40.09 | 62.42 | |
| COFTBase Model=Stable Diffusion v1.5, Training Epochs=202023.06 | 26.92 | — | — | 40.08 | 62.96 | |
| UniControlTraining Setting=Multi-task2024.12 | 25.44 | — | — | — | — | |
| GligenTraining Setting=Single-task2024.12 | 23.78 | — | — | — | — | |
| LoRABase Model=Stable Diffusion v1.5, Training Epochs=202023.06 | 22.98 | — | — | 35.52 | 58.03 | |
| ControlNetBase Model=Stable Diffusion v1.5, Training Epochs=202023.06 | 20.88 | — | — | 30.91 | 61.42 | |
| Uni-ControlNetTraining Setting=Multi-task2024.12 | 19.39 | — | — | — | — | |
| T2I-AdapterBase Model=Stable Diffusion v1.5, Training Epochs=202023.06 | 16.38 | — | — | 26.31 | 51.63 | |
| T2I-AdapterTraining Setting=Single-task2024.12 | 12.61 | — | — | — | — | |
| SDBase Model=Stable Diffusion v1.5, Training Epochs=202023.06 | 7.72 | — | — | 14.4 | 33.61 | |
| NeuTRENO DeiTBackbone=DeiT2023.12 | — | 37.24 | 38.06 | — | — | |
| Softmax DeiTBackbone=DeiT2023.12 | — | 35.72 | 36.68 | — | — |