Semantic Segmentation on NYU V2
63.6mIoUOmniVec2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| OmniVec2Modality=RGBD2025.07 | 63.6 | — | |
| OmniVecModality=RGBD2025.07 | 60.8 | — | |
| GeminiFusionBackbone=Swin-Large-22k, Additional Strategies=false2024.08 | 60.2 | — | |
| StitchFusionBackbone=Swin-Large-22k, Additional Strategies=false2024.08 | 59.6 | — | |
| DPLNetBackbone=MiT-B5, Additional Strategies=true2024.08 | 59.3 | — | |
| EMSANetBackbone=EMSANet-R34-NBt1D, Additional Strategies=true2024.08 | 59 | — | |
| PolyMaXBackbone=ConvNeXt-L, Additional Strategies=true2024.08 | 58.1 | — | |
| SwinMTLBackbone=SwinV2-Base-MiM, Additional Strategies=true2024.08 | 58.1 | — | |
| StitchFusionBackbone=MiT-B2 (-B4,-B5), Additional Strategies=false2024.08 | 57.8 | — | |
| GeminiFusionBackbone=MiT-B5 (MiT-B2), Additional Strategies=false2024.08 | 57.7 | — | |
| DFormerBackbone=DFormer-L, Additional Strategies=true2024.08 | 57.2 | — | |
| CMNeXtBackbone=MiT-B4 (MiT-B2), Additional Strategies=false2024.08 | 56.9 | — | |
| CMXBackbone=MiT-B5, Additional Strategies=false2024.08 | 56.9 | — | |
| Mask3DBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 56.9 | — | |
| OMNIVOREBackbone=Swin-Large, Additional Strategies=true2024.08 | 56.8 | — | |
| OMNIVOREBackbone=Swin-L2022.01 | 56.8 | — | |
| MultiMAEBackbone=ViT-Base, Additional Strategies=true2024.08 | 56 | — | |
| SMMCLBackbone=SegNeXt-B, Additional Strategies=false2024.08 | 55.8 | — | |
| Supervised Pre-trainingBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 55.5 | — | |
| TokenFusionBackbone=MiT-B5 (MiT-B2), Additional Strategies=false2024.08 | 55.1 | — | |
| OMNIVOREBackbone=Swin-B2022.01 | 55.1 | — | |
| MAE-unsupIN→SNBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 54.9 | — | |
| Pri3DBackbone=ResNet-50, Pre-training Data=ImageNet+ScanNet2023.02 | 54.2 | — | |
| StitchFusionBackbone=Swin-Tiny-1k (-22k), Additional Strategies=false2024.08 | 53.8 | — | |
| Pri3DBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 53.2 | — | |
| TCD2022.01 | 53.1 | — | |
| BCMFP + SA-Gate2022.01 | 52.4 | — | |
| MultiMAEInput Modality=RGB, Training Mode=Fine-tuning, Pre-training Dataset=ImageNet-1K2022.04 | 52 | — | |
| Supervised pretrainPre-training dataset=Scenenet2021.07 | 51.6 | — | |
| ShapeConv2022.01 | 51.3 | — | |
| MAEInput Modality=RGB, Training Mode=Fine-tuning, Pre-training Dataset=ImageNet-1K2022.04 | 50.8 | — | |
| Supervised pretrainPre-training dataset=Imagenet2021.07 | 50.3 | — | |
| SupervisedInput Modality=RGB, Training Mode=Fine-tuning, Pre-training Dataset=ImageNet-1K2022.04 | 50.1 | — | |
| ImageNet Pre-training (supIN)Backbone=ResNet-50, Pre-training Data=ImageNet2023.02 | 50 | — | |
| TupleInfoNCE2021.07 | 48.1 | — | |
| DINOInput Modality=RGB, Training Mode=Fine-tuning, Pre-training Dataset=ImageNet-1K2022.04 | 47.9 | — | |
| MoCoV2-supIN→SNBackbone=ResNet-50, Pre-training Data=ImageNet+ScanNet2023.02 | 47.6 | — | |
| MoCo-v3Input Modality=RGB, Training Mode=Fine-tuning, Pre-training Dataset=ImageNet-1K2022.04 | 46.6 | — | |
| MP@Res5backbone=ResNet-50, gating=PAG2018.05 | 46.5 | 73.5 | |
| RefineNetbackbone=ResNet-1522018.05 | 46.5 | 73.6 | |
| MP@Res5backbone=ResNet-50, fusion=w-Avg.2018.05 | 46.3 | 73.4 | |
| MP@Res5backbone=ResNet-50, sparsity (p)=0.92018.05 | 45.8 | 72.9 | |
| PerspectiveParsing2018.05 | 44.5 | 72.1 | |
| RefineNetbackbone=ResNet-502018.05 | 43.8 | — | |
| MP@Res5backbone=ResNet-50, sparsity (p)=0.72018.05 | 43.6 | 71.4 | |
| MISA2021.07 | 43.4 | — | |
| MMV FAC2021.07 | 42.5 | — | |
| baselinebackbone=ResNet-502018.05 | 42.1 | 71.1 | |
| InfoNCE2021.07 | 42.1 | — | |
| CMC2021.07 | 41.9 | — | |
| MP@Res5backbone=ResNet-50, sparsity (p)=0.52018.05 | 40.9 | 69.4 | |
| Train from scratchPre-training=None2021.07 | 40.1 | — | |
| NAYERSynthetic Time=0.82h2023.09 | 38.5 | — | |
| FMSynthetic Time=0.82h2023.09 | 36.6 | — | |
| DFADSynthetic Time=6.0h2023.09 | 36.4 | — | |
| AdaShare# Params=12019.11 | 30.2 | 62.4 | |
| Single-Task# Params=32019.11 | 27.5 | 58.9 | |
| AutoMTL# Params (M)=35.0562021.10 | 26.6 | 58.2 | |
| Single-Task# Params (M)=63.8552021.10 | 26.5 | 58.2 | |
| MTAN# Params (M)=66.2172021.10 | 26 | 57.2 | |
| MTAN# Params=3.112019.11 | 26 | 57.2 | |
| Cross-Stitch# Params (M)=63.8552021.10 | 25.4 | 57.6 | |
| Cross-Stitch# Params=32019.11 | 25.4 | 57.6 | |
| ScratchBackbone=ResNet-50, Pre-training Data=None2023.02 | 24.8 | — | |
| AdaShare# Params (M)=21.2852021.10 | 24.4 | 57.8 | |
| Multi-Task# Params=12019.11 | 24.1 | 57.2 | |
| DEN# Params (M)=23.8382021.10 | 23.9 | 54.9 | |
| DEN# Params=1.122019.11 | 23.9 | 54.9 | |
| Sluice# Params (M)=63.8552021.10 | 23.8 | 56.9 | |
| Sluice# Params=32019.11 | 23.8 | 56.9 | |
| Multi-Task# Params (M)=21.2852021.10 | 22.2 | 54.4 | |
| NDDR-CNN# Params (M)=67.0472021.10 | 21.6 | 53.9 | |
| NDDR-CNN# Params=3.152019.11 | 21.6 | 53.9 | |
| DAFLSynthetic Time=3.99h2023.09 | 10.5 | — |