Research map / Probabilistic and scientific modeling / Molecules and proteins
Human Trajectory Prediction via Counterfactual Analysis
Guangyi Chen, Junlong Li, Jiwen Lu, Jie Zhou
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 counterfactual analysis (zero-vector history intervention, subtract prediction) | STGAT · ETH/UCY (avg of 5 scenes) | ADE/FDE: 0.47/0.93 -> 0.40/0.77 | Improved |
| add counterfactual analysis (zero-vector history intervention) | Social-STGCNN · ETH/UCY (avg of 5 scenes) | ADE/FDE: 0.45/0.75 -> 0.43/0.66 | Improved |
| use mean feature vector instead of zero vector as counterfactual (ablation) | Causal-STGAT · ETH/UCY (avg of 5 scenes) | ADE/FDE: 0.40/0.77 -> 0.44/0.84 | Hurt |
| use random uniform vector instead of zero vector as counterfactual (ablation) | Causal-STGAT · ETH/UCY (avg of 5 scenes) | ADE/FDE: 0.40/0.77 -> 0.42/0.80 | Hurt |
| mean counterfactual versus no counterfactual (ablation) | STGAT · ETH/UCY (avg of 5 scenes) | ADE/FDE: 0.47/0.93 -> 0.44/0.84 | Improved |
| random counterfactual versus no counterfactual (ablation) | STGAT · ETH/UCY (avg of 5 scenes) | ADE/FDE: 0.47/0.93 -> 0.42/0.80 | Improved |
| add causal model to Social-STGCNN with scene image features | S-STGCNN · ETH/UCY with scene image (avg of 5 scenes) | ADE/FDE: 0.57/0.91 -> 0.43/0.73 | Improved |
| add causal model to PECNet | PECNet* · Stanford Drone Dataset | ADE/FDE: 10.27/15.99 -> 9.19/15.31 | Improved |
| add counterfactual analysis | Social-STGCNN* · ETH/UCY (UNIV scene) | ADE/FDE: 0.51/0.93 -> 0.49/0.81 | Improved |
| add counterfactual analysis on ZARA2 scene | Social-STGCNN* · ETH/UCY (ZARA2 scene) | ADE/FDE: 0.31/0.47 -> 0.32/0.49 | Mixed |
| add counterfactual analysis (extra computation on part of network) | Social-STGCNN · ETH/UCY inference | inference time: 0.0116 -> 0.0124 | Hurt |
| add counterfactual analysis | STGAT · ETH/UCY inference | parameters: 56k -> 56k | No clear effect |
Closest papers by meaning
- Causal Intervention for Human Trajectory Prediction with Cross Attention Mechanism (AAAI 2023 · similarity 0.89)
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.87)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.87)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.86)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.86)
- Multi-modal Knowledge Distillation-based Human Trajectory Forecasting (CVPR 2025 · similarity 0.86)
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.
