Research map / Graph learning / Knowledge and temporal graphs
Segmentation-Grounded Scene Graph Generation
Siddhesh Khandelwal, Mohammed Suhail, Leonid Sigal
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 segmentation grounding of objects and relations | MOTIF + VGG-16 · Visual Genome SGCls | mR@100: 9.8 -> 10.5 | Improved |
| add segmentation grounding | MOTIF + ResNeXt-101-FPN · Visual Genome PredCls | mR@100: 19.4 -> 20.2 | Improved |
| add segmentation grounding | VCTree + ResNeXt-101-FPN · Visual Genome SGDet | mR@100: 7.9 -> 9.0 | Improved |
| add segmentation grounding | MOTIF + VGG-16 · Visual Genome SGDet | mR@100: 7.9 -> 8.1 | Improved |
| add segmentation grounding | VCTree + ResNeXt-101-FPN · Visual Genome PredCls zero-shot | zsR@100: 7.1 -> 10.6 | Improved |
| add joint detector pre-training on Visual Genome and COCO (ablation) | VCTree + ResNeXt-101-FPN · Visual Genome SGCls | mR@100: 10.6 -> 11.1 | Improved |
| add object grounding with inferred masks (ablation) | VCTree + Joint · Visual Genome SGCls | mR@100: 11.1 -> 11.8 | Improved |
| add relation (edge) grounding with Gaussian attention (ablation) | VCTree + Joint + OG · Visual Genome SGCls | mR@100: 11.8 -> 12.2 | Improved |
| use Gaussian attention instead of averaging masks for edge grounding (ablation) | VCTree + Joint + OG · Visual Genome SGCls | mR@20: 9.1 -> 9.3 | Improved |
| use Gaussian attention instead of union of masks (ablation) | VCTree + Joint + OG · Visual Genome SGCls | mR@20: 9.1 -> 9.3 | Improved |
| add segmentation mask refinement head (ablation) | VCTree + Joint + OG + EG_Gaussian · Visual Genome SGCls | mR@100: 12.2 -> 12.3 | Improved |
| add segmentation refinement (ablation) | MOTIF + VGG-16 · MSCOCO val PredCls masks | AP: 31.5 -> 42.4 | Improved |
Closest papers by meaning
- Scene Graph Generation by Iterative Message Passing (CVPR 2017 · similarity 0.92)
- Unconditional Scene Graph Generation (ICCV 2021 · similarity 0.91)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.90)
- LinkNet: Relational Embedding for Scene Graph (NeurIPS 2018 · similarity 0.90)
- A Comprehensive Survey of Scene Graphs: Generation and Application (TPAMI 2021 · similarity 0.90)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.90)
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.
