Research map / Graph learning / Knowledge and temporal graphs
Energy-Based Learning for Scene Graph Generation
Check 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 |
|---|---|---|---|
| Train VCTree-TDE with energy-based loss instead of cross-entropy | VCTree-TDE · Visual Genome PredCls | mR@20: 16.3 -> 19.87 | Helped |
| Train VCTree-TDE with energy-based loss instead of cross-entropy | VCTree-TDE · Visual Genome SGCls | mR@20: 11.85 -> 13.86 | Helped |
| Train VCTree-TDE with energy-based loss instead of cross-entropy | VCTree-TDE · Visual Genome SGDet | mR@20: 6.59 -> 7.1 | Helped |
| Train Transformer with energy-based loss instead of cross-entropy | Transformer · GQA PredCls | mR@20: 1.17 -> 1.28 | Helped |
| Train Transformer with energy-based loss instead of cross-entropy | Transformer · GQA SGCls | mR@20: .54 -> .68 | Helped |
| Train VCTree with energy-based loss instead of cross-entropy | VCTree · Visual Genome few-shot relation triplets (1-5 shot) | Recall@20: 16.9 -> 18.55 | Helped |
| Train VCTree with energy-based loss instead of cross-entropy | VCTree · Visual Genome zero-shot PredCls | zsR@20: 1.43 -> 2.25 | Helped |
| Train VCTree with energy-based loss instead of cross-entropy | VCTree · Visual Genome sentence-to-graph retrieval, gallery 1000 | R@20: 14 -> 17.2 | Helped |
| Increase energy optimization from 20 to 120 steps (ablation) | VCTree energy-based training · Visual Genome PredCls | mR@20: 14.2 -> 14.59 | Helped |
| Increase energy optimization from 20 to 40 steps (ablation) | VCTree energy-based training · Visual Genome PredCls | mR@20: 14.2 -> 14.37 | Helped |
| Increase energy optimization from 20 to 60 steps (ablation) | VCTree energy-based training · Visual Genome PredCls | mR@20: 14.2 -> 14.41 | Helped |
| Remove image graph input from energy model (ablation) | VCTree energy-based training · Visual Genome PredCls | mR@20: 14.2 -> 14.19 | Hurt |
Closest papers by meaning
- Towards Unbiased and Robust Spatio-Temporal Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.88)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.88)
- Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency (ICML 2024 · similarity 0.88)
- Semi-Supervised Clustering Framework for Fine-grained Scene Graph Generation (AAAI 2025 · similarity 0.87)
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.87)
- Adaptive Self-training Framework for Fine-grained Scene Graph Generation (ICLR 2024 · similarity 0.87)
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.