Metric: mIoU (higher is better)
| # | Model↕ | mIoU▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Mask2Former (Swin-L) | 47.85 | Yes | Masked-attention Mask Transformer for Universal ... | 2021-12-02 | Code |
| 2 | UPerNet (ConvNeXt-L) | 47.3 | Yes | Unified Perceptual Parsing for Scene Understanding | 2018-07-26 | Code |
| 3 | Mask2Former (ResNet-50) | 43.71 | Yes | Masked-attention Mask Transformer for Universal ... | 2021-12-02 | Code |
| 4 | DeepLabv3 (ResNet-50) | 43.37 | Yes | Rethinking Atrous Convolution for Semantic Image... | 2017-06-17 | Code |
| 5 | Segformer (MiT-B5) | 40.83 | Yes | SegFormer: Simple and Efficient Design for Seman... | 2021-05-31 | Code |