Research map / Detection and segmentation / CNN features
Panoptic-PartFormer: Learning a Unified Model for Panoptic Part Segmentation
Xiangtai Li, Shilin Xu, Yibo Yang, Guangliang Cheng, Yunhai Tong, Dacheng Tao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add the decoupled decoder (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 55.9 -> 57.4 | Improved |
| add dynamic convolution (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 52.2 -> 57.4 | Improved |
| add self attention among queries (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 53.4 -> 57.4 | Improved |
| use 3 interaction stages instead of 1 (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 54.2 -> 57.4 | Improved |
| joint reasoning over thing, stuff and part queries instead of separate reasoning (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 56.8 -> 57.4 | Improved |
| joint reasoning instead of sequential reasoning (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 56.3 -> 57.4 | Improved |
| joint query-based part prediction instead of dense prediction head (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 55.9 -> 57.4 | Improved |
| add ASPP to the dense-prediction part head (ablation) | DP-based baseline · Cityscapes PPS val | PartPQ: 55.9 -> 56.1 | Improved |
| add aligned part decoder (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 56.3 -> 57.4 | Improved |
| also apply aligned decoder to the scene features (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 57.4 -> 57.2 | No clear effect |
| add positional encoding (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 55.1 -> 57.4 | Improved |
| add boundary supervision loss (ablation) | Panoptic-PartFormer · Cityscapes PPS val | PartPQ: 57.4 -> 57.2 | No clear effect |
Later papers that compared with this method
2 later accepted papers report Panoptic-PartFormer as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Panoptic-PartFormer++: A Unified and Decoupled View for Panoptic Part Segmentation (TPAMI 2024 · better on 6 of 6 shared results)
- AIMS: All-Inclusive Multi-Level Segmentation for Anything (NeurIPS 2023 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Panoptic-PartFormer++: A Unified and Decoupled View for Panoptic Part Segmentation (TPAMI 2024 · similarity 0.98)
- Part-Aware Panoptic Segmentation (CVPR 2021 · similarity 0.94)
- Task-aligned Part-aware Panoptic Segmentation through Joint Object-Part Representations (CVPR 2024 · similarity 0.92)
- Unifying Training and Inference for Panoptic Segmentation (CVPR 2020 · similarity 0.91)
- Panoptic Segmentation (CVPR 2019 · similarity 0.91)
- Scene-Centric Unsupervised Panoptic Segmentation (CVPR 2025 · similarity 0.89)
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.
