Research map / Detection and segmentation / CNN features
SOGNet: Scene Overlap Graph Network for Panoptic Segmentation
Read the paperCheck 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 |
|---|---|---|---|
| model and resolve instance overlap relations instead of using heuristic overlap inference | SOGNet · COCO test-dev | PQ: 46.6 -> 47.8 | Helped |
| model and resolve instance overlap relations instead of using heuristic overlap inference | SOGNet · COCO validation | PQ: 42.5 -> 43.7 | Helped |
| model and resolve instance overlap relations instead of using heuristic overlap inference | SOGNet · Cityscapes | PQ: 59.3 -> 60.0 | Helped |
| add relational embedding and differentiable overlap resolution (ablation) | PlainNet + PH1 · COCO validation | PQ: 42.3 -> 43.0 | Helped |
| replace Panoptic Head 1 addition with Panoptic Head 2 gated semantic-mask combination (ablation) | SOGNet · COCO validation | PQ: 43.0 -> 43.7 | Helped |
| add category features to geometry features in relational embedding (ablation) | SOGNet relational embedding · COCO validation (short schedule) | PQ: 37.5 -> 38.3 | Helped |
| add category features to geometry features in relational embedding (ablation) | SOGNet relational embedding · COCO validation (short schedule) | OA: 69.62 -> 88.19 | Helped |
| add geometry features to category features in relational embedding (ablation) | SOGNet relational embedding · COCO validation (short schedule) | PQ: 37.9 -> 38.3 | Helped |
| add appearance (Mask) features to category and geometry features (ablation) | SOGNet relational embedding · COCO validation (short schedule) | OA: 88.19 -> 89.22 | Helped |
| add appearance (Mask) features to category and geometry features (ablation) | SOGNet relational embedding · COCO validation (short schedule) | PQ: 38.3 -> 38.4 | Helped |
| replace the proposed relation supervision with weakly supervised asymmetric overlap labels (ablation) | SOGNet relational embedding · COCO validation (short schedule) | PQ: 38.4 -> 38.4 | No clear effect |
| replace the proposed relation supervision with weakly supervised asymmetric overlap labels (ablation) | SOGNet relational embedding · COCO validation (short schedule) | OA: 89.22 -> 89.31 | No clear effect |
Closest papers by meaning
- Bidirectional Graph Reasoning Network for Panoptic Segmentation (CVPR 2020 · similarity 0.90)
- An End-To-End Network for Panoptic Segmentation (CVPR 2019 · similarity 0.89)
- Unifying Training and Inference for Panoptic Segmentation (CVPR 2020 · similarity 0.89)
- Panoptic, Instance and Semantic Relations: A Relational Context Encoder To Enhance Panoptic Segmentation (CVPR 2022 · similarity 0.89)
- Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · similarity 0.87)
- Task-aligned Part-aware Panoptic Segmentation through Joint Object-Part Representations (CVPR 2024 · similarity 0.85)
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.