Research map / Probabilistic and scientific modeling / Molecules and proteins
It is not the Journey but the Destination: Endpoint Conditioned Trajectory Prediction
Karttikeya Mangalam, Harshayu Girase, Shreyas Agarwal, Kuan-Hui Lee, Ehsan Adeli, Jitendra Malik, Adrien Gaidon
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 truncation trick to endpoint sampling (ablation) | PECNet · Stanford Drone Dataset (K=20) | ADE (pixels): 10.23 -> 9.96 | Improved |
| add truncation trick to endpoint sampling (ablation) | PECNet · Stanford Drone Dataset (K=20) | FDE (pixels): 16.29 -> 15.88 | Improved |
| add non-local social pooling (ablation) | PECNet · Stanford Drone Dataset (K=20) | ADE (pixels): 10.56 -> 10.23 | Improved |
| add non-local social pooling (ablation) | PECNet · Stanford Drone Dataset (K=20) | FDE (pixels): 16.72 -> 16.29 | Improved |
| add social pooling and truncation trick (ablation) | PECNet · ETH/UCY average (K=20) | ADE (m): 0.32 -> 0.29 | Improved |
| add social pooling and truncation trick (ablation) | PECNet · ETH/UCY average (K=20) | FDE (m): 0.54 -> 0.48 | Improved |
| train downstream modules on inferred endpoints instead of ground-truth endpoints (ablation) | PECNet · Stanford Drone Dataset (K=20) | ADE (pixels): 10.87 -> 9.96 | Improved |
| add truncation trick at K=1 (ablation) | PECNet · Stanford Drone Dataset (K=1) | ADE (pixels): 22.85 -> 17.29 | Improved |
| condition on endpoint and use fewer samples | PECNet vs P2TIRL · Stanford Drone Dataset | ADE (pixels): 12.58 -> 12.58 | Improved |
| endpoint conditioning without social pooling or truncation | PECNet · Stanford Drone Dataset (K=20) | ADE (pixels): 12.60 -> 10.56 | Improved |
| endpoint conditioning without social pooling or truncation | PECNet · ETH/UCY average (K=20) | ADE (m): 0.49 -> 0.32 | Improved |
| conditioning on waypoints in the middle of the future instead of endpoint | PECNet · Stanford Drone Dataset | way-point prediction error | Hurt |
Closest papers by meaning
- Trajectron++: Dynamically-Feasible Trajectory Forecasting With Heterogeneous Data (ECCV 2020 · similarity 0.88)
- A Set of Control Points Conditioned Pedestrian Trajectory Prediction (AAAI 2023 · similarity 0.87)
- Social-Mamba: Socially-Aware Trajectory Forecasting with State-Space Models (ECCV 2026 · similarity 0.87)
- From Goals, Waypoints & Paths to Long Term Human Trajectory Forecasting (ICCV 2021 · 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 0.85)
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.
