Glass-like Object Segmentation on Trans10k (test)
90.28mIoUEBLNet
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EBLNetBackbone=ResNet50, Output Stride (OS)=8, Inference Scale=single scale2021.03 | 90.28 | 94.71 | 0.048 | 4.14 | |
| EBLNetBackbone=ResNet50, Output Stride (OS)=16, Inference Scale=single scale2021.03 | 89.58 | 93.95 | 0.052 | 4.6 | |
| TranslabBackbone=ResNet50, Inference Scale=single scale2021.03 | 87.63 | 92.69 | 0.063 | 5.46 | |
| Deeplabv3+Backbone=ResNet50, Inference Scale=single scale2021.03 | 84.54 | 89.54 | 0.081 | 7.78 | |
| Deeplabv3+Backbone=Xception65, Inference Scale=single scale2021.03 | 84.26 | 89.18 | 0.082 | 8 | |
| PSPNetInference Scale=single scale2021.03 | 82.38 | 86.25 | 0.093 | 9.72 | |
| FCNInference Scale=single scale2021.03 | 79.67 | 83.79 | 0.108 | 10.33 | |
| DenseAsppInference Scale=single scale2021.03 | 78.11 | 81.22 | 0.114 | 12.19 | |
| Deeplabv3+Backbone=MobileNetv2, Inference Scale=single scale2021.03 | 75.27 | 80.92 | 0.13 | 12.49 | |
| HRNetInference Scale=single scale2021.03 | 74.56 | 75.82 | 0.134 | 13.52 | |
| BiSeNetInference Scale=single scale2021.03 | 73.93 | 77.92 | 0.14 | 13.96 | |
| RefineNetInference Scale=single scale2021.03 | 66.03 | 57.97 | 0.18 | 22.22 |