Research map / Graph learning / Knowledge and temporal graphs
IS-GGT: Iterative Scene Graph Generation With Generative Transformers
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 |
|---|---|---|---|
| Generate a sparse interaction graph before predicate classification | IS-GGT · Visual Genome, PredCls, SGCls, and SGDet | mR@100 | Mixed |
| Add semantic features to relation prediction (ablation) | IS-GGT · Visual Genome PredCls | mR@100: 28.3 -> 30.1 | Helped |
| Replace GloVe with ConceptNet Numberbatch embeddings (ablation) | IS-GGT · Visual Genome PredCls | mR@100: 30.1 -> 28.5 | No clear effect |
| Add visual features to relation prediction (ablation) | IS-GGT · Visual Genome SGCls | mR@100: 15.2 -> 17.4 | Helped |
| Add global image context to predicate prediction (ablation) | IS-GGT · Visual Genome PredCls | mR@100: 29.3 -> 30.1 | Helped |
| Add graph sampling (ablation) | IS-GGT · Visual Genome PredCls | mR@100: 27.9 -> 30.1 | Helped |
| Use an edge prior to rank sampled edges (ablation) | IS-GGT · Visual Genome SGDet | mR@100: 9.3 -> 11.9 | Helped |
| Use detector-provided node labels instead of predicted node labels (ablation) | IS-GGT · Visual Genome SGDet | mR@100: 8.9 -> 11.9 | Helped |
| Increase the maximum considered edges from 10 to 250 (ablation) | IS-GGT · Visual Genome PredCls | mR@100: 4.6 -> 30.1 | Helped |
| Increase the maximum considered edges from 250 to all edges (ablation) | IS-GGT · Visual Genome SGCls | mR@100: 17.5 -> 17.6 | No clear effect |
Closest papers by meaning
- Iterative Scene Graph Generation (NeurIPS 2022 · similarity 0.92)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.91)
- From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models (CVPR 2024 · similarity 0.90)
- EGTR: Extracting Graph from Transformer for Scene Graph Generation (CVPR 2024 · similarity 0.90)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.89)
- RA-SGG: Retrieval-Augmented Scene Graph Generation Framework via Multi-Prototype Learning (AAAI 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.