Research map / Probabilistic and scientific modeling / Molecules and proteins
Looking to Relations for Future Trajectory Forecast
Chiho Choi, Behzad Dariush
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 2D+3D spatio-temporal encoding instead of 2D+LSTM (ablation) | RE baseline · SDD | ADE/FDE at 4.0 s: 5.81 / 10.52 -> 5.58 / 10.27 | Improved |
| add LSTM temporal modeling after 2D convolutions (ablation) | RE_Conv2D · SDD | ADE/FDE at 4.0 s: 6.19 / 11.92 -> 5.81 / 10.52 | Improved |
| replace 2D-only with 3D convolutions only (ablation) | RE_Conv2D · SDD | ADE/FDE at 4.0 s: 6.19 / 11.92 -> 6.27 / 11.92 | No clear effect |
| add relation gate module (RGM) in place of fully connected layers (ablation) | RE_Conv2D+Conv3D · SDD | ADE/FDE at 4.0 s: 5.58 / 10.27 -> 5.00 / 9.58 | Improved |
| add separate spatial refinement network (ablation) | GRE_Vanilla · SDD | ADE/FDE at 4.0 s: 5.00 / 9.58 -> 4.60 / 8.79 | Improved |
| replace shallow TPN with a deeper single end-to-end network (ablation) | GRE_Vanilla · SDD | ADE/FDE at 4.0 s: 5.00 / 9.58 -> 5.63 / 10.54 | Hurt |
| add interactions of distant regions (ablation) | GRE_Vanilla · SDD | ADE/FDE at 4.0 s: 5.93 / 10.56 -> 5.00 / 9.58 | Improved |
| add MC dropout uncertainty with 2 samples (ablation) | GRE_Refine · SDD | ADE/FDE at 4.0 s: 4.60 / 8.79 -> 4.54 / 8.73 | Improved |
| increase MC dropout samples from 2 to 5 (ablation) | GRE_MC · SDD | ADE/FDE at 4.0 s: 4.54 / 8.73 -> 4.46 / 8.68 | Improved |
| increase MC dropout samples from 5 to 10 (ablation) | GRE_MC · SDD | ADE/FDE at 4.0 s: 4.46 / 8.68 -> 4.47 / 8.65 | No clear effect |
| add spatial refinement network (ablation) | GRE_Vanilla · ETH/UCY average | ADE/FDE: 0.047 / 0.096 -> 0.045 / 0.086 | Improved |
| add MC dropout uncertainty (2 samples) (ablation) | GRE_Refine · ETH/UCY average | ADE/FDE: 0.045 / 0.086 -> 0.044 / 0.086 | Improved |
Closest papers by meaning
- Leveraging Future Relationship Reasoning for Vehicle Trajectory Prediction (ICLR 2023 · similarity 0.88)
- Forecasting Human Trajectory from Scene History (NeurIPS 2022 · similarity 0.88)
- Higher-order Relational Reasoning for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.88)
- Future-Aware Interaction Network For Motion Forecasting (ICCV 2025 · similarity 0.87)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.85)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · 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.
