Research map / Multi-agent systems and games / Human-AI and goal-conditioned
M2I: From Factored Marginal Trajectory Prediction to Interactive Prediction
Qiao Sun, Xin Huang, Junru Gu, Brian C. Williams, Hang Zhao
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 |
|---|---|---|---|
| Factor joint prediction into influencer marginal and reactor conditional predictions | M2I · WOMD interactive validation, All (8s) | mAP: 0.07 -> 0.09 | Improved |
| Factor joint prediction into influencer marginal and reactor conditional predictions | M2I · WOMD interactive test, All (8s) | mAP: 0.06 -> 0.08 | Improved |
| Use a conditional predictor instead of a marginal predictor for the reactor, conditioning on the influencer's ground-truth future (ablation) | M2I · WOMD interactive validation, vehicles (8s) | mAP: 0.30 -> 0.41 | Improved |
| Condition the reactor on only the best predicted influencer trajectory rather than ground truth (ablation) | M2I · WOMD interactive validation, vehicles (8s) | mAP: 0.30 -> 0.26 | Hurt |
| Use TNT within the M2I factorization instead of independent marginal predictions | TNT · WOMD interactive validation, vehicles (8s) | OR: 0.42 -> 0.20 | Improved |
| Use TNT within the M2I factorization instead of independent marginal predictions | TNT · WOMD interactive validation, vehicles (8s) | mAP: 0.10 -> 0.14 | Improved |
| Use the predicted influencer-reactor relations instead of ground-truth relations (ablation) | M2I relation predictor · WOMD validation, vehicles (8s) | mAP: 3.05% | Hurt |
| Train separate marginal and conditional predictors by agent type rather than one model for all agent types (ablation) | M2I · WOMD validation, pedestrians and cyclists | prediction performance | Improved |
Closest papers by meaning
- FJMP: Factorized Joint Multi-Agent Motion Prediction Over Learned Directed Acyclic Interaction Graphs (CVPR 2023 · similarity 0.91)
- Scene Transformer: A unified architecture for predicting future trajectories of multiple agents (ICLR 2022 · similarity 0.91)
- Large Scale Interactive Motion Forecasting for Autonomous Driving: The Waymo Open Motion Dataset (ICCV 2021 · similarity 0.91)
- Density-Adaptive Model Based on Motif Matrix for Multi-Agent Trajectory Prediction (CVPR 2024 · similarity 0.88)
- Trajeglish: Traffic Modeling as Next-Token Prediction (ICLR 2024 · similarity 0.87)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (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.
