Research map / Learning theory and algorithms / Prediction and decision losses
Social-Implicit: Rethinking Trajectory Prediction Evaluation and the Effectiveness of Implicit Maximum Likelihood Estimation
Abduallah Mohamed, Deyao Zhu, Warren Vu, Mohamed Elhoseiny, Christian Claudel
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 Social-Zones (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.16 -> 0.90 | Improved |
| add triplet, geometric distance and angle losses on top of zones (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.07 -> 0.90 | Improved |
| add geometric distance loss to triplet loss with zones (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.03 -> 1.11 | Hurt |
| add geometric angle loss to triplet loss with zones (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.03 -> 0.97 | Improved |
| add triplet loss to Zones (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.07 -> 1.03 | Improved |
| increase number of zones from 1 to 4 (1.2 m/s top zone) (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.16 -> 0.90 | Improved |
| set last zone speed at 0.6 m/s instead of 1.2 m/s (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 0.90 -> 1.19 | Hurt |
| use 2 zones instead of 1 (ablation) | Social-Implicit · ETH/UCY | AMD/AMV: 1.16 -> 1.20 | No clear effect |
| shift predicted samples by 10 cm | Social-Implicit and baselines · ETH/UCY | BoN ADE/FDE | No clear effect |
| evaluate with AMD/AMV instead of ADE/FDE | Trajectron++ vs ExpertTraj · ETH/UCY | AMD: 1.84 -> 38.4 | Mixed |
| train with IMLE (Social-Implicit) vs larger baselines | Social-Implicit · ETH/UCY | parameters: 323.3k -> 5.8K | Improved |
| Social-Implicit vs DAG-Net | Social-Implicit · SDD | KDE: 1.76 -> 3.89 | Hurt |
Closest papers by meaning
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.84)
- SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model (CVPR 2024 · similarity 0.84)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.84)
- Certified Human Trajectory Prediction (CVPR 2025 · similarity 0.83)
- Normalizing Trajectory Models (NeurIPS 2026 · similarity 0.83)
- Evaluation of Trajectory Distribution Predictions with Energy Score (ICML 2024 · similarity 0.83)
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.
