Research map / Multi-agent systems and games / Human-AI and goal-conditioned
SMEMO: Social Memory for Trajectory Forecasting
Francesco Marchetti, Federico Becattini, Lorenzo Seidenari, Alberto Del Bimbo
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 external social memory reading (vs zero reading) (ablation) | SMEMO · SSA (synthetic) | ADE: 0.522 -> 0.169 | Improved |
| add external memory instead of pooled states (ablation) | SMEMO · SSA (synthetic) | ADE: 0.270 -> 0.169 | Improved |
| keep memory across present timestep (vs memory reset) (ablation) | SMEMO · SSA (synthetic) | ADE: 0.180 -> 0.169 | Improved |
| add learned memory reading (vs random reading) (ablation) | SMEMO · SDD (K=20) | ADE: 9.48 -> 8.11 | Improved |
| add learned memory reading (vs zero reading) (ablation) | SMEMO · SDD (K=20) | ADE: 19.30 -> 8.11 | Improved |
| add external memory instead of pooled states (ablation) | SMEMO · SDD (K=20) | ADE: 17.93 -> 8.11 | Improved |
| keep memory across present timestep (vs memory reset) (ablation) | SMEMO · SDD (K=5) | ADE: 12.29 -> 11.64 | Improved |
| use SMEMO write/erase update instead of social-level GRU (ablation) | SMEMO · ETH/UCY average | ADE: 0.24 -> 0.22 | Improved |
| use SMEMO write/erase update instead of social-level GRU (ablation) | SMEMO · SDD (K=20) | ADE: 8.27 -> 8.11 | Improved |
| replace social pooling with SMEMO | PECNet · SDD (K=20) | ADE: 9.96 -> 9.38 | Improved |
| replace social pooling with SMEMO | PECNet · SDD (K=20) | FDE: 15.88 -> 15.15 | Improved |
| add social interaction modelling (PECNet with social pooling vs none) | PECNet · SDD (K=20) | ADE: 10.56 -> 9.96 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AgentFormer — AgentFormer: Agent-Aware Transformers for Socio-Temporal Multi-Agent Forecasting (ICCV 2021) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Social ODE: Multi-agent Trajectory Forecasting with Neural Ordinary Differential Equations (ECCV 2022 · similarity 0.89)
- Social-Mamba: Socially-Aware Trajectory Forecasting with State-Space Models (ECCV 2026 · similarity 0.88)
- Interpretable Social Anchors for Human Trajectory Forecasting in Crowds (CVPR 2021 · similarity 0.87)
- SocialMOIF: Multi-Order Intention Fusion for Pedestrian Trajectory Prediction (CVPR 2025 · similarity 0.86)
- SocialCircle: Learning the Angle-based Social Interaction Representation for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.85)
- 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.
