Research map / Graph learning / Knowledge and temporal graphs
Panoptic Scene Graph Generation
Jingkang Yang, Yi Zhe Ang, Zujin Guo, Kaiyang Zhou, Wayne Zhang, Ziwei Liu
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 |
|---|---|---|---|
| train 60 epochs instead of 12 | PSGTR (ResNet-50) · PSG dataset, scene graph generation (SGDet) | R@100: 4.27 -> 36.3 | Improved |
| train 60 epochs instead of 12 | PSGTR (ResNet-101) · PSG dataset, scene graph generation (SGDet) | R@100: 4.00 -> 35.3 | Improved |
| train 60 epochs instead of 12 | PSGFormer (ResNet-50) · PSG dataset, scene graph generation (SGDet) | R@100: 20.2 -> 20.1 | No clear effect |
| train 60 epochs instead of 12 | PSGFormer (ResNet-50) · PSG dataset, scene graph generation (SGDet) | mR@100: 18.7 -> 17.6 | Hurt |
| model objects and relations with separate decoders and prompting-like matching (PSGFormer vs PSGTR) | PSGTR (12 epochs, ResNet-50) · PSG dataset, scene graph generation (SGDet) | mR@100: 1.57 -> 18.7 | Improved |
| replace VCTree with PSGFormer under 12 epochs | VCTree (ResNet-101) · PSG dataset, scene graph generation (SGDet) | mR@100: 10.3 -> 19.8 | Improved |
| replace VCTree with PSGFormer under 12 epochs | VCTree (ResNet-101) · PSG dataset, scene graph generation (SGDet) | R@100: 23.7 -> 21.0 | No clear effect |
| use GPSNet (explicit predicate direction) instead of MOTIFS | MOTIFS (ResNet-50) · PSG dataset, scene graph generation (SGDet) | R@100: 22.0 -> 20.1 | Hurt |
| use VCTree instead of MOTIFS | MOTIFS (ResNet-50) · PSG dataset, scene graph generation (SGDet) | R@100: 22.0 -> 22.5 | Improved |
| provide ground-truth segmentation (PredCls instead of SGDet) | IMP (ResNet-50) · PSG dataset, R@20 | R@20: 16.5 -> 31.9 | Improved |
| use ResNet-101 instead of ResNet-50 | VCTree · PSG dataset, scene graph generation (SGDet) | R@100: 22.5 -> 23.7 | Improved |
| use ResNet-101 instead of ResNet-50 | PSGTR (12 epochs) · PSG dataset, scene graph generation (SGDet) | R@100: 4.27 -> 4.00 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GPSNet — GPS-Net: Graph Property Sensing Network for Scene Graph Generation (CVPR 2020) · better on 4 of 4 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
- A Fair Ranking and New Model for Panoptic Scene Graph Generation (ECCV 2024 · similarity 0.93)
- Pair Then Relation: Pair-Net for Panoptic Scene Graph Generation (TPAMI 2024 · similarity 0.93)
- OpenPSG: Open-set Panoptic Scene Graph Generation via Large Multimodal Models (ECCV 2024 · similarity 0.92)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.91)
- Panoptic Scene Graph Generation with Semantics-Prototype Learning (AAAI 2024 · similarity 0.90)
- EGTR: Extracting Graph from Transformer for Scene Graph Generation (CVPR 2024 · 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.
