Research map / Video / Action recognition
Weakly-Supervised Action Transition Learning for Stochastic Human Motion 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 |
|---|---|---|---|
| Use weakly supervised action-transition learning rather than omit transition training (ablation) | Ours (RNN) · BABEL | Acc: 15.6^{\pm 0.1} -> 49.6^{\pm 0.4} | Helped |
| Use weakly supervised action-transition learning rather than omit transition training (ablation) | Ours (Transformer) · BABEL | Div: 0.01^{\pm 0.00} -> 1.82^{\pm 0.01} | Helped |
| Add static-pose padding frames (ablation) | Ours (RNN) · GRAB | Acc: 88.4^{\pm 0.6} -> 92.6^{\pm 0.6} | Helped |
| Add static-pose padding frames (ablation) | Ours (RNN) · NTU RGB-D | Acc: 70.1^{\pm 0.2} -> 76.0^{\pm 0.2} | Helped |
| Use weakly supervised pseudo-transitions rather than the limited ground-truth transitions (ablation) | Ours (RNN) · BABEL | Acc: 16.4^{\pm 0.4} -> 49.6^{\pm 0.4} | Helped |
| Use a linear transition-length function instead of an MLP (ablation) | Ours (Transformer) · GRAB | Acc: 65.1^{\pm 1.2} -> 85.5^{\pm 1.2} | Helped |
| Replace padding with a learned stop sign (ablation) | Ours (RNN) · NTU RGB-D | Acc: 29.2^{\pm 0.1} -> 76.0^{\pm 0.2} | Hurt |
| Use action-driven prediction rather than adapted baselines | GRAB · Acc | Acc: 83.0^{\pm 0.3} -> 92.6^{\pm 0.6} | Helped |
| Use action-driven prediction rather than adapted baselines | NTU RGB-D · Acc | Acc: 70.6^{\pm 0.2} -> 76.0^{\pm 0.2} | Helped |
| Use action-driven prediction rather than adapted baselines | BABEL · Acc | Acc: 40.9^{\pm 0.2} -> 49.6^{\pm 0.4} | Helped |
| Use action-driven prediction rather than adapted baselines | HumanAct12 · Acc | Acc: 44.4^{\pm 0.2} -> 59.0^{\pm 0.1} | Helped |
Closest papers by meaning
- Stochastic Human Motion Prediction with Memory of Action Transition and Action Characteristic (CVPR 2025 · similarity 0.92)
- Learning Diverse Stochastic Human-Action Generators by Learning Smooth Latent Transitions (AAAI 2020 · similarity 0.91)
- A Stochastic Conditioning Scheme for Diverse Human Motion Prediction (CVPR 2020 · similarity 0.90)
- MultiAct: Long-Term 3D Human Motion Generation from Multiple Action Labels (AAAI 2023 · similarity 0.89)
- MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation (AAAI 2024 · similarity 0.88)
- FutureHuman3D: Forecasting Complex Long-Term 3D Human Behavior from Video Observations (CVPR 2024 · 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.