Research map / 3D vision / Human pose estimation
Contextually Plausible and Diverse 3D Human Motion Prediction
Sadegh Aliakbarian, Fatemeh Saleh, Lars Petersson, Stephen Gould, Mathieu Salzmann
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 |
|---|---|---|---|
| condition decoder with extended reparameterization on z_c instead of concatenating z_c (ablation) | CVAE decoder · Human3.6M | diversity: 1.28 -> 3.35 | Improved |
| condition decoder with extended reparameterization on z_c instead of concatenating z_c (ablation) | CVAE decoder · Human3.6M | context: 0.40 -> 0.51 | Improved |
| replace deterministic h_t decoder concatenation with stochastic z_c reparameterization (ablation) | CVAE decoder · Human3.6M | diversity: 0.24 -> 3.12 | Improved |
| replace deterministic h_t decoder concatenation with stochastic z_c reparameterization (ablation) | CVAE decoder · Human3.6M | context: 0.43 -> 0.54 | Improved |
| use deterministic h_t instead of stochastic z_c to condition the encoder (ablation) | LCP-VAE encoder · Human3.6M | context: 0.51 -> 0.54 | Improved |
| use deterministic h_t instead of stochastic z_c to condition the encoder (ablation) | LCP-VAE encoder · Human3.6M | diversity: 3.35 -> 3.12 | Hurt |
| learn a conditional prior (LCP-VAE) instead of DLow | stochastic motion prediction · Human3.6M | context: 0.29 -> 0.54 | Improved |
| learn a conditional prior (LCP-VAE) instead of DLow | stochastic motion prediction · Human3.6M | diversity: 4.71 -> 3.12 | Hurt |
| learn a conditional prior (LCP-VAE) instead of Mix-and-Match | stochastic motion prediction · Human3.6M | context: 0.37 -> 0.54 | Improved |
| learn a conditional prior (LCP-VAE) instead of DLow | stochastic motion prediction · CMU MoCap | diversity: 2.90 -> 2.36 | Hurt |
| learn a conditional prior (LCP-VAE) instead of DLow | stochastic motion prediction · Human3.6M (best of K=50 sampled), Walking 80 ms | mean angle error: 0.31 -> 0.22 | Improved |
| LCP-VAE versus Mix-and-Match on long horizon | stochastic motion prediction · Human3.6M (best of K=50 sampled), Walking 1000 ms | mean angle error: 0.68 vs 0.69 | No clear effect |
Closest papers by meaning
- A Stochastic Conditioning Scheme for Diverse Human Motion Prediction (CVPR 2020 · similarity 0.91)
- Generating Smooth Pose Sequences for Diverse Human Motion Prediction (ICCV 2021 · similarity 0.91)
- Harmonizing Stochasticity and Determinism: Scene-responsive Diverse Human Motion Prediction (NeurIPS 2024 · similarity 0.91)
- Motion Diversification Networks (CVPR 2024 · similarity 0.91)
- FutureHuman3D: Forecasting Complex Long-Term 3D Human Behavior from Video Observations (CVPR 2024 · similarity 0.90)
- VideoMDM: Towards 3D Human Motion Generation From 2D Supervision (NeurIPS 2026 · similarity 0.88)
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.
