Research map / 3D vision / Human pose estimation
Another Vertical View: A Hierarchical Network for Heterogeneous Trajectory Prediction via Spectrums
Beihao Xia, Conghao Wong, Duanquan Xu, Qinmu Peng, Xinge You
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 DFT transform to deterministic backbone (ablation) | V2-Net (D) · ETH-UCY (eth) | ADE/FDE: 0.65/1.30 -> 0.62/1.15 | Improved |
| add Haar transform to deterministic backbone (ablation) | V2-Net (D) · ETH-UCY (eth) | ADE/FDE: 0.65/1.30 -> 0.63/1.15 | Improved |
| model dimension interaction (replace SEP separate prediction) (ablation) | V2-Net DFT · ETH-UCY (zara1) | ADE/FDE: 0.21/0.39 -> 0.19/0.30 | Improved |
| add bilinear structure (E-V2-Net) (ablation) | V2-Net DFT · SDD (2D coordinates) | ADE/FDE: 7.12/11.39 -> 6.57/10.50 | Improved |
| model interactions between corner points instead of SEP (ablation) | V2-Net DFT · SDD (2D bounding box) | ADE: 7.06 -> 6.78 | Improved |
| use Haar instead of DFT (ablation) | E-V2-Net · Human3.6M (3D skeleton) | MPJPE at 400ms: 66.62 -> 57.80 | Improved |
| add DFT transform relative to no transform (ablation) | E-V2-Net · Human3.6M (3D skeleton) | MPJPE at 400ms: 64.56 -> 66.62 | Hurt |
| use Haar instead of DFT (ablation) | E-V2-Net · nuScenes (3D bounding box, best-of-20) | ADE/FDE: 0.203/0.291 -> 0.209/0.301 | Hurt |
| use Haar instead of DFT (ablation) | E-V2-Net · SDD (2D bounding box) | ADE: 6.62 -> 6.44 | Improved |
| add transform relative to no transform (ablation) | E-V2-Net · nuScenes (3D bounding box, best-of-20) | ADE/FDE: 0.213/0.308 -> 0.203/0.291 | Improved |
| increase keypoints from 3 to 6 (ablation) | E-V2-Net DFT · SDD (2D bounding box) | ADE: 6.62 -> 6.81 | Hurt |
| use 4 keypoints instead of 1 (ablation) | E-V2-Net Haar · Human3.6M (3D skeleton) | MPJPE at 400ms: 63.58 -> 57.80 | Improved |
Closest papers by meaning
- SimpliHuMoN: Simplifying Human Motion Prediction (NeurIPS 2026 · similarity 0.86)
- Homogeneous Dynamics Space for Heterogeneous Humans (CVPR 2025 · similarity 0.86)
- GigaTraj: Predicting Long-term Trajectories of Hundreds of Pedestrians in Gigapixel Complex Scenes (CVPR 2024 · similarity 0.86)
- SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model (CVPR 2024 · similarity 0.85)
- Multi-agent Long-term 3D Human Pose Forecasting via Interaction-aware Trajectory Conditioning (CVPR 2024 · similarity 0.85)
- MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation (CVPR 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.
