Semantic Segmentation on Chengdu UHR (val)
94.45aAcc (Average Accuracy)GMBFormer
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GMBFormerBackbone=MiT-B42026.06 | 94.45 | 89.25 | 94.25 | 94.31 | 94.38 | 94.25 | |
| SegFormer-B4Backbone=MiT-B42026.06 | 92.99 | 87.4 | 92.71 | 92.83 | 92.98 | 92.71 | |
| Swin-UPerNetBackbone=Swin-Base2026.06 | 92.98 | 87.41 | 92.83 | 92.84 | 92.85 | 92.83 | |
| DeepLabV3Backbone=ResNet-50-D82026.06 | 91.78 | 84.5 | 91.59 | 91.59 | 91.59 | 91.59 | |
| Mask2FormerBackbone=ResNet-502026.06 | 90.98 | 86.12 | 90.74 | 90.77 | 90.81 | 90.74 |