Research map / Graph learning / Knowledge and temporal graphs
HL-Net: Heterophily Learning Network for Scene Graph Generation
Xin Lin, Changxing Ding, Yibing Zhan, Zijian Li, Dacheng Tao
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 |
|---|---|---|---|
| Model heterophilic and homophilic object and relationship context with HL-Net | Scene graph generation · Visual Genome SGCLS | R@100: 41.4 -> 43.5 | Improved |
| Model heterophilic and homophilic object and relationship context with HL-Net | Scene graph generation · Open Images V4 | score_wtd: 46.9 -> 48.1 | Improved |
| Add ART, RFP, and HMP rather than using the ablation baseline (ablation) | HL-Net · Visual Genome SGCLS | R@100: 40.9 -> 43.5 | Improved |
| Add ART to the baseline (ablation) | HL-Net · Visual Genome SGCLS | R@100: 40.9 -> 42.5 | Improved |
| Add RFP to the baseline (ablation) | HL-Net · Visual Genome PREDCLS | R@100: 67.3 -> 68.3 | Improved |
| Add object-only HMP to the baseline (ablation) | HL-Net · Visual Genome SGCLS | R@100: 40.9 -> 41.9 | Improved |
| Add HMP to the ART-and-RFP variant (ablation) | HL-Net · Visual Genome PREDCLS | R@100: 68.5 -> 68.9 | Improved |
| Set tau to 0.5 rather than 0.2 or 0.7 (ablation) | HL-Net ART · Visual Genome SGCLS | R@50: 42.3, 42.6, 42.4 | Improved |
| Increase ART depth from 2-step to 5-step (ablation) | HL-Net ART · Visual Genome SGCLS | R@100: 42.8 -> 43.5 | Improved |
| Use 4-step rather than 5-step RFP (ablation) | HL-Net RFP · Visual Genome PREDCLS | R@100: 68.6 -> 68.9 | Improved |
Later papers that compared with this method
2 later accepted papers report HL-Net as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · better on 0 of 2 shared results)
- Visual Traffic Knowledge Graph Generation from Scene Images (ICCV 2023 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unbiased Heterogeneous Scene Graph Generation with Relation-Aware Message Passing Neural Network (AAAI 2023 · similarity 0.93)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.89)
- Kumaraswamy Wavelet for Heterophilic Scene Graph Generation (AAAI 2024 · similarity 0.89)
- EGTR: Extracting Graph from Transformer for Scene Graph Generation (CVPR 2024 · similarity 0.89)
- Scene Graph Generation Strategy with Co-occurrence Knowledge and Learnable Term Frequency (ICML 2024 · similarity 0.89)
- Scene Graph-Grounded Image 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.
