Command Palette
Search for a command to run...
Panoptic Segmentation On Cityscapes Test
评估指标
PQ
评测结果
各个模型在此基准测试上的表现结果
| Paper Title | ||
|---|---|---|
| OneFormer (ConvNeXt-L, single-scale, Mapillary Vistas-Pretrained) | 68.0 | OneFormer: One Transformer to Rule Universal Image Segmentation |
| Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary, multi-scale) | 67.8 | Scaling Wide Residual Networks for Panoptic Segmentation |
| EfficientPS | 67.1 | EfficientPS: Efficient Panoptic Segmentation |
| Axial-DeepLab-XL (Mapillary Vistas, multi-scale) | 66.6 | Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation |
| kMaX-DeepLab (single-scale) | 66.2 | kMaX-DeepLab: k-means Mask Transformer |
| Panoptic-Deeplab | 65.5 | Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation |
| EfficientPS (Cityscapes-fine) | 62.9 | EfficientPS: Efficient Panoptic Segmentation |
| SOGNet (ResNet-50) | 60 | SOGNet: Scene Overlap Graph Network for Panoptic Segmentation |
| COPS (ResNet-50) | 60 | Combinatorial Optimization for Panoptic Segmentation: A Fully Differentiable Approach |
| Dynamically Instantiated Network | 55.4 | Pixelwise Instance Segmentation with a Dynamically Instantiated Network |
0 of 10 row(s) selected.