Research map / Detection and segmentation / CNN features
Improving Panoptic Segmentation at All Scales
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Train on crops rather than full images (ablation) | HRNetV2-W48+ panoptic segmentation · Mapillary Vistas | PQ: 39.4 -> 43.6 | Helped |
| Train on crops rather than full images (ablation) | HRNetV2-W48+ panoptic segmentation · Indian Driving Dataset | PQ: 49.1 -> 50.3 | Helped |
| Train on crops rather than full images (ablation) | HRNetV2-W48+ panoptic segmentation · Cityscapes | mAP: 39.5 -> 42.2 | Helped |
| Add CABB loss to crop training (ablation) | HRNetV2-W48+ panoptic segmentation · Mapillary Vistas | PQ: 43.6 -> 44.5 | Helped |
| Add ISUS sampling to crop training (ablation) | HRNetV2-W48+ panoptic segmentation · Mapillary Vistas | PQ: 43.6 -> 44.7 | Helped |
| Add CABB loss and ISUS together (ablation) | HRNetV2-W48+ crop-trained panoptic segmentation · Mapillary Vistas | PQ: 43.6 -> 45.1 | Helped |
| Add CABB loss and ISUS together (ablation) | HRNetV2-W48+ crop-trained panoptic segmentation · Indian Driving Dataset | PQ: 50.3 -> 50.7 | Helped |
| Add CABB loss and ISUS together (ablation) | HRNetV2-W48+ crop-trained panoptic segmentation · Cityscapes | mAP: 42.2 -> 43.4 | Helped |
| Use instance-aware scale sampling rather than uniform scale augmentation in the same range (ablation) | Crop-trained HRNetV2-W48+ · Mapillary Vistas | PQ-th: 42.3 -> 43.1 | Helped |
| Use instance-aware scale sampling rather than uniform scale augmentation in the same range (ablation) | Crop-trained HRNetV2-W48+ · Mapillary Vistas | mAP: 22.8 -> 23.0 | Helped |
| Add CABB to crop training (ablation) | Crop-trained panoptic segmentation · Mapillary Vistas, largest objects | mAP | Helped |
| Add ISUS to crop training (ablation) | Crop-trained panoptic segmentation · Mapillary Vistas, smallest objects | mAP | Hurt |
Closest papers by meaning
- Unifying Training and Inference for Panoptic Segmentation (CVPR 2020 · similarity 0.88)
- PanopticDepth: A Unified Framework for Depth-Aware Panoptic Segmentation (CVPR 2022 · similarity 0.87)
- Real-Time Panoptic Segmentation From Dense Detections (CVPR 2020 · similarity 0.87)
- LPSNet: A Lightweight Solution for Fast Panoptic Segmentation (CVPR 2021 · similarity 0.87)
- Ada-Segment: Automated Multi-loss Adaptation for Panoptic Segmentation (AAAI 2021 · similarity 0.87)
- Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.