Research map / Graph learning / Knowledge and temporal graphs
NODIS: Neural Ordinary Differential Scene Understanding
Yuren Cong, Hanno Ackermann, Wentong Liao, Michael Ying Yang, Bodo Rosenhahn
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 |
|---|---|---|---|
| add object ODE (O-ODE) (ablation) | ours-LSTM without ODEs · Visual Genome SGCLS (graph constraints) | R@50: 35.3 -> 41.4 | Improved |
| add object ODE (O-ODE) when P-ODE present (ablation) | ours-LSTM · Visual Genome SGCLS (graph constraints) | R@50: 35.9 -> 41.9 | Improved |
| add predicate ODE (P-ODE) (ablation) | ours-LSTM without ODEs · Visual Genome SGCLS (graph constraints) | R@50: 35.3 -> 35.9 | Improved |
| add predicate ODE (P-ODE) when O-ODE present (ablation) | ours-LSTM · Visual Genome SGCLS (graph constraints) | R@50: 41.4 -> 41.9 | Improved |
| add predicate ODE (P-ODE) (ablation) | ours-LSTM · Visual Genome PREDCLS (graph constraints) | R@50: 65.4 -> 65.9 | No clear effect |
| add both ODE modules (ablation) | ours-LSTM · Visual Genome SGGEN (graph constraints) | R@50: 27.1 -> 27.7 | Improved |
| add O-ODE (ablation) | ours-LSTM · Visual Genome PREDCLS (graph constraints) | R@50: 65.4 -> 65.4 | No clear effect |
| use linear-layer function (FC) instead of GCNN front-end | ours · Visual Genome SGCLS (graph constraints) | R@50: 38.2 -> 41.7 | Improved |
| use bidirectional LSTM instead of FC front-end | ours-FC · Visual Genome SGCLS (graph constraints) | R@50: 41.7 -> 41.9 | No clear effect |
| increase ODE integral time t_end from 0.05 to 1.5 (ablation) | ours-LSTM · Visual Genome SGCLS (graph constraints) | R@50: 56.5 -> 62.0 | Improved |
| increase ODE integral time t_end beyond 1.5 (to 3.00) (ablation) | ours-LSTM · Visual Genome SGCLS (graph constraints) | R@50 | Hurt |
| use ODE module in place of CMAT for SGCLS | ours-LSTM · Visual Genome SGCLS (graph constraints) | R@100: 39.8 -> 42.9 | Improved |
Closest papers by meaning
- LinkNet: Relational Embedding for Scene Graph (NeurIPS 2018 · similarity 0.87)
- A Comprehensive Survey of Scene Graphs: Generation and Application (TPAMI 2021 · similarity 0.87)
- Scene Graph Generation by Iterative Message Passing (CVPR 2017 · similarity 0.87)
- Classification by Attention: Scene Graph Classification with Prior Knowledge (AAAI 2021 · similarity 0.87)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.85)
- Noise-Guided Predicate Representation Extraction and Diffusion-Enhanced Discretization for Scene Graph Generation (ICML 2025 · similarity 0.85)
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.
