Cuboid layout estimation on Stanford 2D-3D (test)
87.243D IoUOurs
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OursBackbone=HRNet-18, repro_codebase=PanoLayoutStudio2023.11 | 87.24 | — | — | |
| Dula-Netv2Training Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=false2023.03 | 86.6 | 2.48 | 0.67 | |
| DuLaNet v2Backbone=ResNet-502023.11 | 86.6 | — | — | |
| LGT-NetTraining Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=true2023.03 | 86.03 | 2.11 | 0.63 | |
| LGT-NetBackbone=ResNet-502023.11 | 86.03 | — | — | |
| LGT-NetBackbone=HRNet-18, repro_codebase=PanoLayoutStudio2023.11 | 85.83 | — | — | |
| LGT-NetTraining Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=false2023.03 | 85.76 | — | — | |
| DOPNetTraining Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=true2023.03 | 85.58 | 2.1 | 0.66 | |
| DOPNetTraining Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=false2023.03 | 85.47 | — | — | |
| DOPNetTraining Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=false2023.03 | 85.46 | — | — | |
| LGT-NetTraining Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=false2023.03 | 85.16 | — | — | |
| LayoutNetv2Training Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=false2023.03 | 85.02 | 1.79 | 0.63 | |
| DOPNetTraining Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=true2023.03 | 85 | 2.13 | 0.69 | |
| LGT-NetTraining Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=true2023.03 | 84.94 | 2.07 | 0.69 | |
| SSLayout360Number of labels=916, Number of images=9492021.03 | 84.66 | 1.97 | 0.6 | |
| ResNet-50Backbone=ResNet-50, Pre-training=ImageNet, para.=82M2023.08 | 84.66 | 2.04 | 66 | |
| PanoSwinTBackbone=PanoSwinT, Pre-training=ImageNet, para.=46M2023.08 | 84.21 | 1.98 | 65 | |
| PanoSwinT92Backbone=PanoSwinT92, Pre-training=ImageNet, para.=44M2023.08 | 84.11 | 2 | 65 | |
| Swin-TBackbone=Swin-T, Pre-training=ImageNet, para.=44M2023.08 | 84.04 | 2.07 | 66 | |
| AtlantaNetTraining Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=false2023.03 | 83.94 | 2.18 | 0.71 | |
| AtlantaNetBackbone=ResNet-502023.11 | 83.94 | — | — | |
| ResNet-34Backbone=ResNet-34, Pre-training=ImageNet, para.=33M2023.08 | 83.88 | 2.14 | 68 | |
| Dula-Netv2Training Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=false2023.03 | 83.77 | 2.43 | 0.81 | |
| OmniLayoutTraining Set=PanoContext + Stanford 2D-3D, Input Resolution=512 x 10242021.04 | 83.4 | 2.14 | 0.68 | |
| HorizonNetNumber of labels=916, Number of images=9492021.03 | 82.79 | 2.13 | 0.64 | |
| HorizonNetTraining Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=false2023.03 | 82.72 | 2.27 | 0.69 | |
| HorizonNetBackbone=ResNet-502023.11 | 82.72 | — | — | |
| LayoutNetv2Training Configuration=Stnfd.2D3D + Whole PanoContext, Post-processing=false2023.03 | 82.66 | 2.59 | 0.83 | |
| LayoutNet v2Backbone=ResNet-342023.11 | 82.66 | — | — | |
| HorizonNetTraining Configuration=PanoContext + Whole Stnfd.2D3D, Post-processing=false2023.03 | 82.63 | 2.17 | 0.74 | |
| OmniLayoutTraining set=Stanford 3D-3D, Input resolution=512 x 10242021.04 | 81.2 | 2.37 | 0.78 | |
| HorizonNetTraining Set=PanoContext + Stanford 2D-3D, Input Resolution=512 x 10242021.04 | 80.8 | 2.36 | 0.77 | |
| SSLayout360Number of labels=200, Number of images=9492021.03 | 79.78 | 2.91 | 1.04 | |
| DuLa-NetTraining set=Stanford 3D-3D, Input resolution=256 x 5122021.04 | 79.63 | — | — | |
| ResNet-50Backbone=ResNet-50, Pre-training=None, para.=82M2023.08 | 78.12 | 2.91 | 90 | |
| PanoSwinT92Backbone=PanoSwinT92, Pre-training=None, para.=44M2023.08 | 78.1 | 2.99 | 92 | |
| Swin-TBackbone=Swin-T, Pre-training=None, para.=44M2023.08 | 78 | 3.05 | 94 | |
| ResNet-34Backbone=ResNet-34, Pre-training=None, para.=33M2023.08 | 77.86 | 3.01 | 92 | |
| ResNet-50 + SphereConvBackbone=ResNet-50, Module=SphereConv, Pre-training=None, para.=82M2023.08 | 77.64 | 2.94 | 90 | |
| LayoutNetTraining Set=PanoContext + Stanford 2D-3D, Input Resolution=512 x 10242021.04 | 77.51 | 2.42 | 0.92 | |
| HorizonNetTraining set=Stanford 3D-3D, Input resolution=512 x 10242021.04 | 77.2 | 2.5 | 0.97 | |
| SSLayout360Number of labels=100, Number of images=9492021.03 | 76.96 | 3.01 | 1.15 | |
| LayoutNetTraining set=Stanford 3D-3D, Input resolution=512 x 10242021.04 | 76.33 | 2.7 | 1.04 | |
| HorizonNetNumber of labels=200, Number of images=9492021.03 | 74.95 | 3.69 | 1.5 | |
| SSLayout360Number of labels=50, Number of images=9492021.03 | 73.86 | 3.24 | 1.32 | |
| SSLayout360Number of labels=20, Number of images=9492021.03 | 71.6 | 3.5 | 1.69 | |
| HorizonNetNumber of labels=100, Number of images=9492021.03 | 69.94 | 3.77 | 1.66 | |
| HorizonNetNumber of labels=50, Number of images=9492021.03 | 68.27 | 3.95 | 1.64 | |
| HorizonNetNumber of labels=20, Number of images=9492021.03 | 62.2 | 5.03 | 2.7 |