Research map / Vision-language and multimodal / Human and embodied modeling
Linguistic Structures As Weak Supervision for Visual Scene Graph Generation
Keren Ye, Adriana Kovashka
Check 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 phrasal context to the Basic model (ablation) | Scene graph generation · VG-GT-Graph, both VG splits; VG-Cap-Graph, both splits | R@50: 2.20 -> 2.77; 3.82 -> 4.04; 0.81 -> 0.90; 0.99 -> 1.39 | Improved |
| Add phrasal context to the Basic model (ablation) | Scene graph generation · COCO-Cap-Graph, both splits | R@50: 1.20 -> 1.17; 2.09 -> 1.65 | Hurt |
| Add iterative detection refinement after phrasal context (ablation) | Scene graph generation · VG-GT-Graph and Cap-Graph, all reported splits | R@50: 2.77 -> 3.26; 4.04 -> 6.06; 0.90 -> 1.11; 1.39 -> 1.79; 1.17 -> 1.41; 1.65 -> 2.41 | Improved |
| Add sequential context using an RNN to rerank and relabel triplets (ablation) | Scene graph generation · VG-GT-Graph and Cap-Graph, all reported splits | R@50: 3.26 -> 4.92; 6.06 -> 7.30; 1.11 -> 1.83; 1.79 -> 3.85; 1.41 -> 1.95; 2.41 -> 3.28 | Improved |
| Use caption-derived text graphs instead of ground-truth text graphs | Scene graph generation · Visual Genome scene graph evaluation | R@50: 4.92 vs 1.83; 7.30 vs 3.85 | Hurt |
Closest papers by meaning
- LLM4SGG: Large Language Models for Weakly Supervised Scene Graph Generation (CVPR 2024 · similarity 0.91)
- Scene Graph Generation with Role-Playing Large Language Models (NeurIPS 2024 · similarity 0.89)
- A Closer Look at Dynamic Scene Graph Generation in the Era of Multimodal Large Language Models (NeurIPS 2026 · similarity 0.89)
- Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · similarity 0.89)
- The Scene Language: Representing Scenes with Programs, Words, and Embeddings (CVPR 2025 · similarity 0.88)
- Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation (AAAI 2025 · similarity 0.88)
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.
