Semantic Segmentation on NYUD v2 (test)
60.9mIoUGeminiFusion*
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GeminiFusion*Backbone=Swin-Large-22k, Additional Strategies=true, Pre-training=SUN RGBD2024.06 | 60.9 | — | — | — | — | — | |
| OmniVecBackbone=OmniVec-4, Additional Strategies=true2024.06 | 60.8 | — | — | — | — | — | |
| GeminiFusionBackbone=Swin-Large-22k, Additional Strategies=false2024.06 | 60.2 | — | — | — | — | — | |
| DPLNetBackbone=MiT-B5, Params (M)=7.15, Scale=MS2023.12 | 59.3 | — | — | — | — | — | |
| DPLNetBackbone=MiT-B5, Additional Strategies=false2024.06 | 59.3 | — | — | — | — | — | |
| EMSANetBackbone=EMSANet-R34-NBt1D, Additional Strategies=false2024.06 | 59 | — | — | — | — | — | |
| DPLNetBackbone=MiT-B5, Params (M)=7.15, Scale=SS2023.12 | 58.3 | — | — | — | — | — | |
| PolyMaXBackbone=ConvNeXt-L, Additional Strategies=true2024.06 | 58.1 | — | — | — | — | — | |
| SwinMTLBackbone=SwinV2-Base-MiM, Additional Strategies=true2024.06 | 58.1 | — | — | — | — | — | |
| GeminiFusionBackbone=MiT-B5, Additional Strategies=false2024.06 | 57.7 | — | — | — | — | — | |
| DFormer-LBackbone=DFormer-L, Params (M)=39.0, Scale=MS2023.12 | 57.2 | — | — | — | — | — | |
| DFormerBackbone=DFormer-L, Additional Strategies=true2024.06 | 57.2 | — | — | — | — | — | |
| CMX-B5Backbone=MiT-B5, Params (M)=181.1, Scale=MS2023.12 | 56.9 | — | — | — | — | — | |
| CMXNeXtBackbone=MiT-B4, Params (M)=119.6, Scale=MS2023.12 | 56.9 | — | — | — | — | — | |
| CMNeXtBackbone=MiT-B4, Additional Strategies=false2024.06 | 56.9 | — | — | — | — | — | |
| CMXBackbone=MiT-B5, Additional Strategies=false2024.06 | 56.9 | — | — | — | — | — | |
| OMNIVOREBackbone=Swin-Large, Additional Strategies=true2024.06 | 56.8 | — | — | — | — | — | |
| GeminiFusionBackbone=MiT-B3, Additional Strategies=false2024.06 | 56.8 | — | — | — | — | — | |
| MultiMAEBackbone=ViT-B, Params (M)=95.2, Scale=SS2023.12 | 56 | — | — | — | — | — | |
| MultiMAEBackbone=ViT-Base, Additional Strategies=true2024.06 | 56 | — | — | — | — | — | |
| SMMCLBackbone=SegNeXt-B, Additional Strategies=false2024.06 | 55.8 | — | — | — | — | — | |
| TokenFusion†Backbone=MiT-B5, Additional Strategies=false, Reproduced=true2024.06 | 55.1 | — | — | — | — | — | |
| TokenFusion-B3Backbone=MiT-B3, Params (M)=45.9, Scale=SS2023.12 | 54.2 | — | — | — | — | — | |
| Omnivore-BBackbone=Swin-B, Params (M)=95.7, Scale=SS2023.12 | 54 | — | — | — | — | — | |
| PGDENetBackbone=ResNet-34, Params (M)=100.7, Scale=SS2023.12 | 53.7 | — | — | — | — | — | |
| InvPT2022.03 | 53.56 | — | — | — | — | — | |
| HS3-FuseBackbone=ResNet-101, Multi-Scale Inference=true2021.11 | 53.5 | 78.3 | — | — | — | — | |
| DCANetMulti-scale testing=true2022.10 | 53.3 | 78.2 | — | — | — | — | |
| InverseFormBackbone=ResNet-101, Multi-Scale Inference=true2021.11 | 53.1 | 78.1 | — | — | — | — | |
| InverseFormMulti-scale testing=true2022.10 | 53.1 | 78.1 | — | — | — | — | |
| CEN-PSPNetBackbone=ResNet-152, Multi-Scale Inference=true2021.11 | 52.5 | 77.7 | — | — | — | — | |
| Ours-PSPNetModality=RGB-D, Backbone Network=ResNet1522021.12 | 52.5 | 77.7 | 65 | — | — | — | |
| SA-Gate2021.05 | 52.4 | 77.9 | — | — | — | — | |
| SA-GateBackbone=ResNet-101, Multi-Scale Inference=true2021.11 | 52.4 | 77.9 | — | — | — | — | |
| SA-GateMulti-scale testing=true2022.10 | 52.4 | 77.9 | — | — | — | — | |
| SA-GateBackbone=ResNet-101, Params (M)=110.9, Scale=MS2023.12 | 52.4 | — | — | — | — | — | |
| NANetBackbone=ResNet-101, Multi-Scale Inference=true2021.11 | 52.3 | 77.9 | — | — | — | — | |
| HS3-FuseBackbone=ResNet-101, Multi-Scale Inference=false2021.11 | 52.2 | 77.4 | — | — | — | — | |
| Ours-RefineNetModality=RGB-D, Backbone Network=ResNet152, Evaluation Scale=multi-scale2021.12 | 52.2 | 77.4 | 64.8 | — | — | — | |
| GeminiFusionBackbone=Swin-Tiny-1k, Additional Strategies=false2024.06 | 52.2 | — | — | — | — | — | |
| DCANetMulti-scale testing=false2022.10 | 52.1 | 77.3 | — | — | — | — | |
| FSFNetBackbone=ResNet-1012021.05 | 52 | 77.9 | — | — | — | — | |
| FSFNetBackbone=ResNet-101, Additional Strategies=false2024.06 | 52 | — | — | — | — | — | |
| InverseFormBackbone=ResNet-101, Multi-Scale Inference=false2021.11 | 51.9 | 77.1 | — | — | — | — | |
| Ours-RefineNetModality=RGB-D, Backbone Network=ResNet101, Evaluation Scale=multi-scale2021.12 | 51.7 | 77.2 | 63.7 | — | — | — | |
| Ours-RefineNetModality=RGB-D, Backbone Network=ResNet152, Evaluation Scale=single-scale2021.12 | 51.6 | 77 | 64.4 | — | — | — | |
| SA-GateBackbone=ResNet-101, Multi-Scale Inference=false2021.11 | 51.5 | 76.8 | — | — | — | — | |
| ShapeConvMulti-scale testing=true2022.10 | 51.3 | 76.4 | — | — | — | — | |
| AsymFusionData modality=RGB-D, Backbone=ResNet152, #Params.=133.89M2021.08 | 51.2 | 77 | 64 | — | — | — | |
| SGNet*Backbone=ResNet101, MS (Multi-scale test)=true, SI (Spatial Information)=Depth, FPS=26, param (M)=64.7, ASPP=true2020.04 | 51.1 | 76.8 | 63.3 | — | — | — | |
| CEN-RefineNetBackbone=ResNet-152, Multi-Scale Inference=false2021.11 | 51.1 | — | — | — | — | — | |
| Ours-RefineNetModality=RGB-D, Backbone Network=ResNet101, Evaluation Scale=single-scale2021.12 | 51.1 | 76.2 | 62.8 | — | — | — | |
| SGNetMulti-scale testing=true2022.10 | 51.1 | 76.8 | — | — | — | — | |
| SGNetBackbone=ResNet-101, Params (M)=64.7, Scale=MS2023.12 | 51.1 | — | — | — | — | — | |
| PSDBackbone=ResNet50, Additional Strategies=false2024.06 | 51 | — | — | — | — | — | |
| Malleable 2.5DBackbone=ResNet-101, Multi-Scale Inference=true2021.11 | 50.9 | 76.9 | — | — | — | — | |
| M2.5DMulti-scale testing=true2022.10 | 50.9 | 76.9 | — | — | — | — | |
| AsymFusionData modality=RGB-D, Backbone=ResNet101, #Params.=118.20M2021.08 | 50.8 | 76.6 | 63.5 | — | — | — | |
| PAP2021.05 | 50.4 | 76.2 | — | — | — | — | |
| SGNet*Backbone=ResNet101, MS (Multi-scale test)=false, SI (Spatial Information)=Depth, FPS=26, param (M)=64.7, ASPP=true2020.04 | 50.2 | 76.1 | 62.7 | — | — | — | |
| ShapeConvMulti-scale testing=false2022.10 | 50.2 | 75.8 | — | — | — | — | |
| RDFNetBackbone=2xResNet152, MS (Multi-scale test)=true, SI (Spatial Information)=HHA, FPS=9, param (M)=200.12020.04 | 50.1 | 76 | 62.8 | — | — | — | |
| RDFNetModality=RGB-D, Backbone Network=ResNet1522021.12 | 50.1 | 76 | 62.8 | — | — | — | |
| RDFMulti-scale testing=true2022.10 | 50.1 | 76 | — | — | — | — | |
| RDFNetData modality=RGB-D, Backbone=ResNet152, #Params.=398.00M2021.08 | 50.1 | 76 | 62.8 | — | — | — | |
| SGNetBackbone=ResNet101, MS (Multi-scale test)=false, SI (Spatial Information)=Depth, FPS=28, param (M)=58.32020.04 | 49.6 | 75.6 | 61.9 | — | — | — | |
| SSMAModality=RGB-D, Backbone Network=ResNet101, re-implemented=true2021.12 | 49.6 | 75.8 | 62.3 | — | — | — | |
| SCNModality=RGB-D, Backbone Network=ResNet1522021.12 | 49.6 | — | — | — | — | — | |
| SCNData modality=RGB-D, Backbone=ResNet1522021.08 | 49.6 | — | — | — | — | — | |
| RDFNetBackbone=2xResNet101, MS (Multi-scale test)=true, SI (Spatial Information)=HHA, FPS=11, param (M)=169.12020.04 | 49.1 | 75.6 | 62.2 | — | — | — | |
| RDFNetModality=RGB-D, Backbone Network=ResNet1012021.12 | 49.1 | 75.6 | 62.2 | — | — | — | |
| RDFNetData modality=RGB-D, Backbone=ResNet101, #Params.=366.71M2021.08 | 49.1 | 75.6 | 62.2 | — | — | — | |
| CBNModality=RGB-D, Backbone Network=ResNet101, re-implemented=true2021.12 | 48.9 | 75.5 | 61.2 | — | — | — | |
| SSMAModality=RGB-D, Backbone Network=ResNet502021.12 | 48.7 | 75.2 | 60.5 | — | — | — | |
| D-CNNBackbone=2xResNet152, MS (Multi-scale test)=false, SI (Spatial Information)=Depth2020.04 | 48.4 | — | 61.1 | — | — | — | |
| D-CNNMulti-scale testing=false2022.10 | 48.4 | — | — | — | — | — | |
| ACNetBackbone=2xResNet50, MS (Multi-scale test)=false, SI (Spatial Information)=Depth, FPS=18, param (M)=116.62020.04 | 48.3 | — | — | — | — | — | |
| ACNet2021.05 | 48.3 | — | — | — | — | — | |
| ACNetMulti-scale testing=false2022.10 | 48.3 | — | — | — | — | — | |
| ACNetBackbone=ResNet-50, Params (M)=116.6, Scale=SS2023.12 | 48.3 | — | — | — | — | — | |
| ACNetModality=RGB-D, Backbone Network=ResNet502021.12 | 48.1 | — | 48.3 | — | — | — | |
| CFNetBackbone=2xResNet152, MS (Multi-scale test)=true, SI (Spatial Information)=HHA2020.04 | 47.7 | — | — | — | — | — | |
| SGNetBackbone=ResNet50, MS (Multi-scale test)=false, SI (Spatial Information)=Depth, FPS=39, param (M)=39.32020.04 | 47.7 | 75 | 59.6 | — | — | — | |
| CFN2021.05 | 47.7 | — | — | — | — | — | |
| CFNModality=RGB-D, Backbone Network=ResNet1522021.12 | 47.7 | — | — | — | — | — | |
| CFNData modality=RGB-D, Backbone=ResNet1522021.08 | 47.7 | — | — | — | — | — | |
| RefineNetModality=RGB, Backbone Network=ResNet1522021.12 | 47.6 | 74.4 | 59.6 | — | — | — | |
| RefineNetData modality=RGB, Backbone=ResNet152, #Params.=133.74M2021.08 | 47.6 | 74.4 | 59.6 | — | — | — | |
| RefineNetBackbone=ResNet152, MS (Multi-scale test)=true, FPS=16, param (M)=129.52020.04 | 46.5 | 73.6 | 58.9 | — | — | — | |
| RefineNetData modality=RGB, Backbone=ResNet101, #Params.=118.10M, implementation=re-implemented2021.08 | 46.5 | 73.8 | 59 | — | — | — | |
| RefineNetModality=RGB, Backbone Network=ResNet1012021.12 | 46.4 | 73.8 | 58.8 | — | — | — | |
| RefineNetData modality=RGB, Backbone=ResNet101, #Params.=118.10M2021.08 | 46.4 | 73.8 | 58.8 | — | — | — | |
| ATRC2022.03 | 46.33 | — | — | — | — | — | |
| MTI-Net2022.03 | 45.97 | — | — | — | — | — | |
| LSD-GFBackbone=2xVGG16, MS (Multi-scale test)=false, SI (Spatial Information)=HHA2020.04 | 45.9 | 71.9 | 60.7 | — | — | — | |
| LSD-GFMulti-scale testing=false2022.10 | 45.9 | 71.9 | — | — | — | — | |
| MMAF-NetMulti-scale testing=false2022.10 | 44.8 | 72.2 | — | — | — | — | |
| Kong et al.2021.05 | 44.5 | 72.1 | — | — | — | — | |
| FedPerBackbone=Swin-T, Client Setup=Multi-Task2023.11 | 44.02 | — | -1.11 | — | — | — | |
| Single Task - DITASKTrainable Params (in M)=1.612025.02 | 44.01 | — | 18.04 | — | — | — |