Research map / Graph learning / Knowledge and temporal graphs
Disentangled Multi-Relational Graph Convolutional Network for Pedestrian Trajectory Prediction
Inhwan Bae, Hae-Gon Jeon
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 |
|---|---|---|---|
| use DMRGCN instead of Social-STGCNN | pedestrian trajectory prediction · AVG (ETH and UCY) | ADE: 0.44 -> 0.34 | Improved |
| use DMRGCN instead of Social-STGCNN | pedestrian trajectory prediction · AVG (ETH and UCY) | FDE: 0.75 -> 0.58 | Improved |
| add global temporal aggregation to the GCN-based model (ablation) | DMRGCN · AVG | FDE: 0.63 -> 0.58 | Improved |
| add global temporal aggregation without GCN (ablation) | DMRGCN variant with no GCN · AVG | FDE: 0.74 -> 0.72 | No clear effect |
| use disentangled multi-scale aggregation rather than ordinary multi-scale aggregation (ablation) | DMRGCN · AVG | FDE: 0.69 -> 0.68 | Improved |
| add both distance and relative-displacement relations rather than distance alone (ablation) | DMRGCN with no aggregation · AVG | FDE: 0.71 -> 0.70 | Improved |
| add both distance and relative-displacement relations rather than displacement alone (ablation) | DMRGCN with no aggregation · AVG | ADE: 0.39 -> 0.38 | No clear effect |
| vary DropEdge rate across 0.9, 0.8, 0.7, 0.6, and 0.5 (ablation) | DMRGCN · AVG | ADE: 0.37, 0.35, 0.37, 0.37, 0.37 | Mixed |
| vary the number of TPCNN blocks from 1 to 6 (ablation) | DMRGCN with one GCN and DropEdge 0.8 · AVG | FDE: 0.69, 0.70, 0.68, 0.63, 0.70, 0.73 | Mixed |
| increase the number of GCNs from one to two (ablation) | DMRGCN with GTA · AVG | FDE: 0.58 -> 0.61 | Hurt |
| vary distance-relation scale set (ablation) | DMRGCN · AVG | FDE: 0.72, 0.71, 0.69, 0.72, 0.70 | Mixed |
| vary relative-displacement scale set (ablation) | DMRGCN · AVG | FDE: 0.73, 0.71, 0.70, 0.69, 0.70 | Mixed |
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.
- Social-STGCNN — Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction (CVPR 2020) · better on 8 of 8 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
- A Set of Control Points Conditioned Pedestrian Trajectory Prediction (AAAI 2023 · similarity 0.93)
- Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction (CVPR 2020 · similarity 0.92)
- A Unified Environmental Network for Pedestrian Trajectory Prediction (AAAI 2024 · similarity 0.89)
- Higher-order Relational Reasoning for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.89)
- MS-TIP: Imputation Aware Pedestrian Trajectory Prediction (ICML 2024 · similarity 0.87)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.87)
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.
