Research map / Vision-language and multimodal / Human and embodied modeling
VLPrompt-PSG: Vision-Language Prompting for Panoptic Scene Graph Generation
Zijian Zhou, Holger Caesar, Qijun Chen, Miaojing Shi
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 language information from LLMs (ablation) | VLPrompt · PSG (SGDet) | mR@100: 44.2 -> 53.7 | Improved |
| add language features for rare relations | VLPrompt · PSG rare relations | mR@100: 37.9 -> 51.7 | Improved |
| add language features for common relations | VLPrompt · PSG common relations | mR@100: 54.7 -> 57.0 | Improved |
| add chain-of-thought prompts (ablation) | VLPrompt · PSG (SGDet) | mR@100: 45.7 -> 53.7 | Improved |
| take object features from pixel decoder rather than transformer decoder (ablation) | VLPrompt · PSG (SGDet) | mR@100: 50.3 -> 53.7 | Improved |
| use mask pooling rather than bbox pooling (ablation) | VLPrompt · PSG (SGDet) | mR@100: 52.6 -> 53.7 | Improved |
| concatenate object features rather than subtract (ablation) | VLPrompt · PSG (SGDet) | mR@100: 48.9 -> 53.7 | Improved |
| add spatial features (ablation) | VLPrompt · PSG (SGDet) | mR@100: 52.1 -> 53.7 | Improved |
| use separate RP and RJ feature extraction instead of swapping (ablation) | VLPrompt · PSG (SGDet) | mR@100: 46.4 -> 53.7 | Improved |
| use both RP and RJ decoders instead of RP only (ablation) | VLPrompt · PSG (SGDet) | mR@100: 49.6 -> 53.7 | Improved |
| use 2 decoder blocks instead of 1 (ablation) | VLPrompt · PSG (SGDet) | mR@100: 46.9 -> 53.7 | Improved |
| use 12 decoder blocks instead of 2 (ablation) | VLPrompt · PSG (SGDet) | mR@100: 53.7 -> 29.4 | Hurt |
Closest papers by meaning
- From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models (CVPR 2024 · similarity 0.92)
- Panoptic Scene Graph Generation with Semantics-Prototype Learning (AAAI 2024 · similarity 0.91)
- Pair Then Relation: Pair-Net for Panoptic Scene Graph Generation (TPAMI 2024 · similarity 0.90)
- Panoptic Scene Graph Generation (ECCV 2022 · similarity 0.90)
- Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation (AAAI 2025 · similarity 0.90)
- DSGG: Dense Relation Transformer for an End-to-end 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.
