Research map / Architectures and efficiency / RNNs and normalization
SR-LSTM: State Refinement for LSTM Towards Pedestrian Trajectory Prediction
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SR-LSTM state refinement with two iterations, current-neighbor states and both selection modules | V-LSTM · ETH/UCY AVG | MAD: 0.53 -> 0.45 | Helped |
| add motion gate to simple refinement (ablation) | SR-LSTM · ETH/UCY AVG | MAD: 0.52 -> 0.48 | Helped |
| add pedestrian-wise attention to simple refinement (ablation) | SR-LSTM · ETH/UCY AVG | MAD: 0.52 -> 0.49 | Helped |
| add both motion gate and pedestrian-wise attention (ablation) | SR-LSTM simple refinement · ETH/UCY AVG | FAD: 1.16 -> 0.97 | Helped |
| increase neighborhood size from 2 to 10 with simple, unselected message passing (ablation) | SR-LSTM simple refinement · ETH/UCY AVG | MAD: 0.50 -> 0.52 | Hurt |
| increase neighborhood size from 2 to 10 with both selection modules (ablation) | SR-LSTM · ETH/UCY AVG | MAD: 0.49 -> 0.46 | Helped |
| use current rather than previous neighboring hidden states (ablation) | SR-LSTM · ETH/UCY AVG | MAD: 0.49 -> 0.46 | Helped |
| add a second refinement iteration (ablation) | SR-LSTM · ETH/UCY AVG | FAD: 0.97 -> 0.94 | Helped |
| add a third refinement iteration after two iterations (ablation) | SR-LSTM · ETH/UCY AVG | MAD: 0.45 -> 0.45 | No clear effect |
| replace relative position with normalized absolute position (ablation) | V-LSTM · ETH/UCY AVG | MAD: 0.76 -> 0.63 | Helped |
| correct ETH-univ frame rate (ablation) | V-LSTM · ETH/UCY AVG | MAD: 0.63 -> 0.55 | Helped |
| add random rotation to frame-rate-corrected preprocessing (ablation) | V-LSTM · ETH/UCY AVG | MAD: 0.55 -> 0.53 | Helped |
Closest papers by meaning
- Social LSTM: Human Trajectory Prediction in Crowded Spaces (CVPR 2016 · similarity 0.92)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.88)
- SocialMOIF: Multi-Order Intention Fusion for Pedestrian Trajectory Prediction (CVPR 2025 · similarity 0.88)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.87)
- Higher-order Relational Reasoning for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.86)
- A Unified Environmental Network for Pedestrian Trajectory Prediction (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.