Research map / Video / Video segmentation and tracking
Towards Scene Graph Anticipation
Rohith Peddi, Saksham Singh, Saurabh ., Parag Singla, Vibhav Gogate
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 temporal encoder and dual decoding of observed and anticipated relationships (baseline+ to baseline++) (ablation) | STTran / DSGDetr anticipation baselines · Action Genome AGS, 30% observed, With Constraint | R@10: 12.5 -> 18.5 (STTran), 12.8 -> 19.5 (DSGDetr) | Improved |
| replace anticipatory transformer with NeuralODE latent dynamics (SceneSayerODE) | anticipation baselines · Action Genome AGS, 30% observed, With Constraint | R@10: 19.5 -> 23.1 | Improved |
| use NeuralSDE instead of NeuralODE for latent dynamics (ablation) | SceneSayer · Action Genome AGS, 30% observed, With Constraint | R@10: 23.1 -> 25.0 | Improved |
| use SceneSayerSDE on No Constraint R@50 | SceneSayer · Action Genome AGS, 30% observed, No Constraint | R@50: 48.9 -> 47.4 | Hurt |
| use SceneSayerSDE instead of best baseline | scene graph anticipation · Action Genome AGS | R@10: up to ~70% better | Improved |
| use SceneSayerSDE instead of best baseline | scene graph anticipation · Action Genome PGAGS | R@10: up to ~30% improvement | Improved |
| use SceneSayerSDE instead of best baseline | scene graph anticipation · Action Genome GAGS | R@10: approximately 26% improvement | Improved |
| use SceneSayer ODE/SDE for mean recall | scene graph anticipation · Action Genome AGS, 30% observed, With Constraint | mR@10: 5.9 -> 11.4 | Improved |
| freeze pre-trained ORPU and SCPU and train only LDPU instead of end-to-end training (ablation) | SceneSayer · Action Genome | Recall (qualitative) | Hurt |
| remove objects violating continuity from the test set (ablation) | all methods · Action Genome GAGS, 30% observed, No Constraint | R@10: 46.9 -> 70.9 (SceneSayerSDE) | Improved |
| provide more scene information (AGS to PGAGS to GAGS) (ablation) | SceneSayerSDE · Action Genome, 30% observed, With Constraint | R@10: 25.0 -> 28.8 -> 39.7 | Improved |
Closest papers by meaning
- Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (NeurIPS 2026 · similarity 0.90)
- Towards Unbiased and Robust Spatio-Temporal Scene Graph Generation and Anticipation (CVPR 2025 · similarity 0.90)
- A Closer Look at Dynamic Scene Graph Generation in the Era of Multimodal Large Language Models (NeurIPS 2026 · similarity 0.88)
- Adaptive Visual Scene Understanding: Incremental Scene Graph Generation (NeurIPS 2024 · similarity 0.87)
- DSGG: Dense Relation Transformer for an End-to-end Scene Graph Generation (CVPR 2024 · similarity 0.87)
- OED: Towards One-stage End-to-End Dynamic Scene Graph Generation (CVPR 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.
