Semantic Segmentation on Stanford-2D-3D
83.7IoUMP@Res5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MP@Res5backbone=ResNet-50, fusion=w-Avg.2018.05 | 83.7 | 93.6 | |
| MP@Res5backbone=ResNet-50, gating=PAG2018.05 | 83.7 | 93.7 | |
| MP@Res5backbone=ResNet-50, sparsity (p)=0.92018.05 | 82.8 | 93.3 | |
| MP@Res5backbone=ResNet-50, sparsity (p)=0.72018.05 | 82.4 | 93.2 | |
| MP@Res5backbone=ResNet-50, sparsity (p)=0.52018.05 | 81.8 | 92.9 | |
| baselinebackbone=ResNet-502018.05 | 79.5 | 92.1 | |
| loop2 (test-aug)Backbone=ResNet-101, Number of loops=2, Depth Usage=predicted depth, test-time augmentation=true2017.05 | 76.5 | 91 | |
| PerspectiveParsing2018.05 | 76.5 | 91 | |
| loop2Backbone=ResNet-101, Number of loops=2, Depth Usage=predicted depth2017.05 | 76 | 90.8 | |
| loop1 w/ pred-depthBackbone=ResNet-101, Number of loops=1, Depth Usage=predicted depth2017.05 | 75.3 | 90.6 | |
| loop1 w/ gt-depthBackbone=ResNet-101, Number of loops=1, Depth Usage=ground-truth depth2017.05 | 74.7 | 90.2 | |
| loop1 w/o depthBackbone=ResNet-101, Number of loops=1, Depth Usage=none2017.05 | 74.4 | 90.1 | |
| w/ pred-depthBackbone=ResNet-101, Depth Usage=predicted depth2017.05 | 74.2 | 90 | |
| w/ gt-depthBackbone=ResNet-101, Depth Usage=ground-truth depth2017.05 | 73 | 89.7 | |
| DeepLab2017.05 | 69.8 | 88 | |
| DeepLabnotes=re-trained model2018.05 | 69.8 | 88 | |
| PSPNetevaluation protocol=trained by authors with released code2017.05 | 67.4 | 87.6 | |
| PSPNetnotes=re-trained model2018.05 | 67.4 | 87.6 | |
| baselineBackbone=ResNet-1012017.05 | 64.4 | 86.6 | |
| R2R-U-Netgrid resolution=r = 5, architecture=Residual U-Net2019.02 | 39.4 | 55.9 | |
| Jiang et al.grid resolution=r = 52019.02 | 38.3 | 54.7 |