Metric: PQth (higher is better)
| # | Model↕ | PQth▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Panoptic FCN* (Swin-L, single-scale) | 40.8 | No | Fully Convolutional Networks for Panoptic Segmen... | 2020-12-01 | Code |
| 2 | OneFormer (ConvNeXt-L, single-scale) | 40.6 | No | OneFormer: One Transformer to Rule Universal Ima... | 2022-11-10 | Code |
| 3 | OneFormer (DiNAT-L, single-scale) | 40.5 | No | OneFormer: One Transformer to Rule Universal Ima... | 2022-11-10 | Code |
| 4 | Panoptic-DeepLab (SWideRNet-(1, 1, 4.5), multi-scale) | 39.3 | No | Scaling Wide Residual Networks for Panoptic Segm... | 2020-11-23 | - |
| 5 | Mask2Former + Intra-Batch Supervision (ResNet-50) | 34.9 | No | Intra-Batch Supervision for Panoptic Segmentatio... | 2023-04-17 | Code |
| 6 | Axial-DeepLab-L (multi-scale) | 33.4 | No | Axial-DeepLab: Stand-Alone Axial-Attention for P... | 2020-03-17 | Code |
| 7 | Panoptic FCN* (ResNet-FPN) | 32.9 | No | Fully Convolutional Networks for Panoptic Segmen... | 2020-12-01 | Code |