Research map / Graph learning / Knowledge and temporal graphs
Spatial-Temporal Consistency Network for Low-Latency Trajectory Forecasting
Shijie Li, Yanying Zhou, Jinhui Yi, Juergen Gall
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 |
|---|---|---|---|
| use L2 on observed and NLL on forecast instead of NLL on both (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.63 / 1.09 -> 0.51 / 0.85 | Improved |
| use NLL instead of L2 on the forecast trajectory (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.93 / 1.79 -> 0.51 / 0.85 | Improved |
| weight w of 0.7 or 1.3 instead of 1.0 on the loss terms (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.51 / 0.85 -> 0.54 / 0.87 | Hurt |
| concatenate reconstructed observed trajectory before refinement (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.58 / 0.97 -> 0.51 / 0.85 | Improved |
| add final refinement dilated temporal convolutions (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.60 / 1.04 -> 0.51 / 0.85 | Improved |
| add graph convolutions (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.64 / 1.02 -> 0.51 / 0.85 | Improved |
| add dilated temporal convolutions (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.59 / 0.99 -> 0.51 / 0.85 | Improved |
| use feature-wise convolution instead of TXP-CNN (ablation) | STC-Net · Stanford Drone (world) | ADE / FDE: 0.55 / 0.91 -> 0.51 / 0.85 | Improved |
| search architecture with ENAS (STC-Net-NAS) (ablation) | STC-Net · Stanford Drone (image) | ADE / FDE: 11.96 / 20.12 -> 11.88 / 20.08 | Improved |
| add 3 more layers (L=6) (ablation) | STC-Net · Stanford Drone (image) | ADE / FDE: 11.96 / 20.12 -> 11.61 / 18.79 | Improved |
| widen channels 4x (ablation) | STC-Net · Stanford Drone (image) | ADE / FDE: 11.96 / 20.12 -> 11.49 / 19.04 | Improved |
| combine depth L=6 and 4x channels (ablation) | STC-Net · Stanford Drone (image) | ADE / FDE: 11.96 / 20.12 -> 11.34 / 18.65 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- STGAT — STGAT: Modeling Spatial-Temporal Interactions for Human Trajectory Prediction (ICCV 2019) · better on 4 of 4 shared results
- Social-STGCNN — Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction (CVPR 2020) · better on 2 of 2 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
- Implicit Latent Variable Model for Scene-Consistent Motion Forecasting (ECCV 2020 · similarity 0.88)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · similarity 0.88)
- STraj: Self-training for Bridging the Cross-Geography Gap in Trajectory Prediction (AAAI 2025 · similarity 0.86)
- AutoTraces: Autoregressive Trajectory Forecasting via Multimodal Large Language Models (CVPR 2026 · similarity 0.86)
- Trajectory Mamba: Efficient Attention-Mamba Forecasting Model Based on Selective SSM (CVPR 2025 · similarity 0.86)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · 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.
