Semantic Segmentation on SUN-RGBD (test)
53.3mIoUGeminiFusion24
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GeminiFusion24Backbone=MiT-B5, Params=137.2M, Input size=530 x 730, Flops=332.4G2025.04 | 53.3 | — | — | — | |
| DFormerv2-LBackbone=DFormerv2-Large, Params=95.5M, Input size=530 x 730, Flops=160.5G2025.04 | 53.3 | — | — | — | |
| DFormerv2-BBackbone=DFormerv2-Base, Params=53.9M, Input size=530 x 730, Flops=86.9G2025.04 | 52.8 | — | — | — | |
| GeminiFusion24Backbone=MiT-B3, Params=75.8M, Input size=530 x 730, Flops=179.0G2025.04 | 52.7 | — | — | — | |
| DFormer24Backbone=DFormer-Large, Params=39.0M, Input size=530 x 730, Flops=84.5G2025.04 | 52.5 | — | — | — | |
| CMXBackbone=MiT-B5, Multi-scale test=true2022.03 | 52.4 | 83.8 | — | — | |
| CMX22Backbone=MiT-B5, Params=181.1M, Input size=530 x 730, Flops=217.6G2025.04 | 52.4 | — | — | — | |
| CMXBackbone=MiT-B4, Multi-scale test=true2022.03 | 52.1 | 83.5 | — | — | |
| CMX22Backbone=MiT-B4, Params=139.9M, Input size=530 x 730, Flops=173.8G2025.04 | 52.1 | — | — | — | |
| CMNext23Backbone=MiT-B4, Params=119.6M, Input size=530 x 730, Flops=170.3G2025.04 | 51.9 | — | — | — | |
| FRNet22Backbone=ResNet-34, Params=85.5M, Input size=530 x 730, Flops=150.0G2025.04 | 51.8 | — | — | — | |
| DFormerv2-SBackbone=DFormerv2-Small, Params=26.7M, Input size=530 x 730, Flops=43.7G2025.04 | 51.5 | — | — | — | |
| TokenFusion22Backbone=MiT-B3, Params=45.9M, Input size=530 x 730, Flops=122.1G2025.04 | 51.4 | — | — | — | |
| DFormer24Backbone=DFormer-Base, Params=29.5M, Input size=530 x 730, Flops=54.0G2025.04 | 51.2 | — | — | — | |
| CEN20Backbone=ResNet-152, Params=133.9M, Input size=530 x 730, Flops=849.7G2025.04 | 51.1 | — | — | — | |
| MultiMAE22Backbone=ViT-Base, Params=95.2M, Input size=640 x 640, Flops=267.9G2025.04 | 51.1 | — | — | — | |
| PGDENet22Backbone=ResNet-34, Params=100.7M, Input size=530 x 730, Flops=229.1G2025.04 | 51 | — | — | — | |
| PAP2020.07 | 50.5 | 83.8 | — | — | |
| TokenFusion22Backbone=MiT-B2, Params=26.0M, Input size=530 x 730, Flops=71.1G2025.04 | 50.3 | — | — | — | |
| CEN20Backbone=ResNet-101, Params=118.2M, Input size=530 x 730, Flops=790.3G2025.04 | 50.2 | — | — | — | |
| DFormer24Backbone=DFormer-Small, Params=18.7M, Input size=530 x 730, Flops=33.0G2025.04 | 50 | — | — | — | |
| CMXBackbone=MiT-B2, Multi-scale test=true2022.03 | 49.7 | 82.8 | — | — | |
| CMX22Backbone=MiT-B2, Params=66.6M, Input size=530 x 730, Flops=86.3G2025.04 | 49.7 | — | — | — | |
| TCD2022.03 | 49.5 | 83.1 | — | — | |
| SA-Gatemodality=RGB-D2020.07 | 49.4 | 82.5 | — | — | |
| SA-Gate2022.03 | 49.4 | 82.5 | — | — | |
| SA-Gate20Backbone=ResNet-101, Params=110.9M, Input size=530 x 730, Flops=250.1G2025.04 | 49.4 | — | — | — | |
| AsymFormer24Backbone=MiT-B0+ConvNeXt-Tiny, Params=33.0M, Input size=530 x 730, Flops=52.6G2025.04 | 49.1 | — | — | — | |
| NANet2022.03 | 48.8 | 82.3 | — | — | |
| DFormer24Backbone=DFormer-Tiny, Params=6.0M, Input size=530 x 730, Flops=15.0G2025.04 | 48.8 | — | — | — | |
| SGNet2022.03 | 48.6 | 82 | — | — | |
| ShapeConv2022.03 | 48.6 | 82.2 | — | — | |
| SGNet20Backbone=ResNet-101, Params=64.7M, Input size=530 x 730, Flops=151.2G2025.04 | 48.6 | — | — | — | |
| ShapeConv21Backbone=ResNext-101, Params=86.8M, Input size=530 x 730, Flops=161.8G2025.04 | 48.6 | — | — | — | |
| SGNet-101Backbone=ResNet-101, Multi-scale=true2021.08 | 48.6 | 82 | 60.7 | — | |
| ShapeConv (Ours-101*)Backbone=ResNet-101, Multi-scale=true2021.08 | 48.6 | 82.2 | 59.2 | 71.3 | |
| EMSANet22Backbone=ResNet-34, Params=46.9M, Input size=530 x 730, Flops=58.6G2025.04 | 48.4 | — | — | — | |
| CANetModality=RGB, Number of classes=372020.02 | 48.3 | 82 | — | — | |
| CFN2020.07 | 48.1 | — | — | — | |
| ACNet2020.07 | 48.1 | — | — | — | |
| CFN2022.03 | 48.1 | — | — | — | |
| ACNet2022.03 | 48.1 | — | — | — | |
| CFN-101Backbone=ResNet-101, Multi-scale=true2021.08 | 48.1 | — | — | — | |
| RDF-1522020.07 | 47.7 | 81.5 | — | — | |
| RDF-152Backbone=ResNet-1522022.03 | 47.7 | 81.5 | — | — | |
| RDF-152Backbone=ResNet-152, Multi-scale=true2021.08 | 47.7 | 81.5 | 60.1 | — | |
| ShapeConv (Ours-101)Backbone=ResNet-1012021.08 | 47.6 | 82 | 58.5 | 71.2 | |
| RGBD baselinemodality=RGB-D2020.07 | 47.5 | 81.5 | — | — | |
| SGNet-101Backbone=ResNet-1012021.08 | 47.5 | 81 | 59.8 | — | |
| Ding et al.Modality=RGB, Number of classes=372020.02 | 47.1 | 81.4 | — | — | |
| MMAF-Net-152Backbone=ResNet-1522021.08 | 47 | 81 | 58.2 | — | |
| RGB baselinemodality=RGB2020.07 | 46 | 81 | — | — | |
| 3DGNN2020.07 | 45.9 | — | — | — | |
| 3DGNN2022.03 | 45.9 | — | — | — | |
| 3DGNN-101Backbone=ResNet-101, Multi-scale=true2021.08 | 45.9 | — | 57 | — | |
| RefineNet-101Data=RGB2021.07 | 45.7 | 80.4 | — | — | |
| RefineNetModality=RGB, Number of classes=372020.02 | 45.7 | 80.4 | — | — | |
| Kong et al.2020.07 | 45.1 | 80.3 | — | — | |
| CI-NetData=RGB2021.07 | 44.3 | 80.7 | — | — | |
| 3DGNN-101Backbone=ResNet-1012021.08 | 44.1 | — | 55.7 | — | |
| SSMAData=RGBD2021.07 | 43.9 | 80.2 | — | — | |
| ContextData=RGB2021.07 | 42.3 | 78.4 | — | — | |
| PiecewiseModality=RGB, Number of classes=372020.02 | 42.3 | 78.4 | — | — | |
| Depth-aware CNN2020.07 | 42 | — | — | — | |
| D-CNN2022.03 | 42 | — | — | — | |
| D-CNN-50Backbone=ResNet-502021.08 | 42 | — | 53.5 | — | |
| D-CNNData=RGBD2021.07 | 42 | — | — | — | |
| CMoDEData=RGBD2021.07 | 41.9 | 79.8 | — | — | |
| LFCData=RGBD2021.07 | 41.8 | 79.4 | — | — | |
| AdaptNet++Data=RGB2021.07 | 38.4 | — | — | — | |
| FuseNetData=RGBD2021.07 | 37.3 | 76.3 | — | — | |
| DeepLabv2Modality=RGB, Number of classes=372020.02 | 32.1 | 71.9 | — | — | |
| SegNetModality=RGB, Number of classes=372020.02 | 31.8 | 72.6 | — | — | |
| B-SegNetData=RGB2021.07 | 30.7 | 71.2 | — | — | |
| Bayesian SegNetModality=RGB, Number of classes=372020.02 | 30.7 | 71.2 | — | — | |
| FCNModality=RGB, Number of classes=372020.02 | 27.4 | 68.2 | — | — | |
| DeconvNetModality=RGB, Number of classes=372020.02 | 22.6 | 66.1 | — | — |