Research map / Graph learning / Knowledge and temporal graphs
Fine-Grained Scene Graph Generation via Sample-Level Bias Prediction
Yansheng Li, Tingzhu Wang, Kang Wu, Linlin Wang, Xin Guo, Wenbin Wang
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 |
|---|---|---|---|
| predict from union-region context instead of entire image (ablation) | SBG (Transformer) · VG PredCls | A@50: 40.3 -> 44.6 | Improved |
| add global bias to union-feature bias predictor (ablation) | SBG (Transformer) · VG PredCls | A@50: 42.6 -> 44.6 | Improved |
| add union features to global-bias-only predictor (ablation) | SBG (Transformer) · VG PredCls | A@50: 42.1 -> 44.6 | Improved |
| use global bias only without union features (ablation) | Transformer · VG PredCls | A@50: 41.9 -> 42.1 | No clear effect |
| apply sample-level bias correction (SBG) instead of dataset-level DLFE | Motif PredCls · VG | A@100: 41.5 -> 45.9 | Improved |
| apply SBG instead of dataset-level RTPB | Motif PredCls · VG | mR@100: 37.4 vs 34.4 mR@100; A@100 40.0 -> 45.9 | Mixed |
| add SBG to a classic SGG model | Motif · VG PredCls | R@100: 67.2 -> 57.3 | Hurt |
| add SBG to a classic SGG model | Motif · VG PredCls | mR@100: 19.3 -> 34.4 | Improved |
| apply SBG on GQA (smaller dataset, weaker long tail) | Motif · GQA PredCls | A@50: 47.2 -> 47.4 vs CFA | No clear effect |
| apply SBG on VG-1800 (severe long tail) | Motif · VG-1800 PredCls | F-Acc Top-1: 5.33 -> 13.45 | Improved |
| apply SBP to Faster R-CNN on COCO | Faster R-CNN R-50-FPN 1x · COCO | mAP: 36.4 -> 37.6 | Improved |
| apply SBG to one-stage SGTR | SGTR · VG SGDet | A@100: FGPL-A 23.8 vs SBG 23.6; A@50 20.2 vs 18.6 | Mixed |
Closest papers by meaning
- Foundation Models Based Scene Graph Generation (TPAMI 2026 · similarity 0.91)
- Panoptic Scene Graph Generation with Semantics-Prototype Learning (AAAI 2024 · similarity 0.90)
- TD²-Net: Toward Denoising and Debiasing for Video Scene Graph Generation (AAAI 2024 · similarity 0.89)
- Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency (ICML 2024 · similarity 0.89)
- Towards Unbiased and Robust Spatio-Temporal Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.89)
- Semi-Supervised Clustering Framework for Fine-grained Scene Graph Generation (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.
