Research map / 3D vision / Human pose estimation
Spatiotemporal Co-Attention Recurrent Neural Networks for Human-Skeleton Motion Prediction
Xiangbo Shu, Liyan Zhang, Guo-Jun Qi, Wei Liu, Jinhui Tang
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 skeleton-joint co-attention (SCA) (ablation) | SC-RNN · H3.6M avg 15 actions, 1000ms | mean angle error: 1.52 -> 1.46 | Improved |
| add skeleton-joint co-attention (SCA) (ablation) | SC-RNN · H3.6M avg 15 actions, 80ms | mean angle error: 0.49 -> 0.37 | Improved |
| add skeleton-joint co-attention (SCA) (ablation) | SC-RNN · H3.6M avg 15 actions, 400ms | mean angle error: 1.20 -> 1.04 | Improved |
| add joint attention to skeleton attention (ablation) | SC-RNN · H3.6M avg 15 actions, 1000ms | mean angle error: 1.51 -> 1.46 | Improved |
| add skeleton attention to joint attention (ablation) | SC-RNN · H3.6M avg 15 actions, 1000ms | mean angle error: 1.49 -> 1.46 | Improved |
| add joint attention to skeleton attention at short horizon (ablation) | SC-RNN · H3.6M avg 15 actions, 400ms | mean angle error: 1.06 -> 1.04 | No clear effect |
| use traveling-based joint sequence instead of ID-based sequence (ablation) | SC-RNN · H3.6M, 1000ms | mean angle error: 1.50 -> 1.46 | Improved |
| use traveling-based joint sequence instead of surrounding-based sequence (ablation) | SC-RNN · H3.6M, 1000ms | mean angle error: 1.48 -> 1.46 | Improved |
| use traveling-based joint sequence instead of ID-based sequence (ablation) | SC-RNN · H3.6M, 80ms | mean angle error: 0.40 -> 0.37 | Improved |
| use SC-RNN instead of MHU | SC-RNN · H3.6M avg 15 actions, 1000ms | mean angle error: 1.54 -> 1.46 | Improved |
| use SC-RNN instead of MHU at 80ms | SC-RNN · H3.6M avg 15 actions, 80ms | mean angle error: 0.37 -> 0.37 | No clear effect |
| use SC-RNN instead of MHU on greeting action | SC-RNN · H3.6M greeting, 1000ms | mean angle error: 1.88 -> 1.64 | Improved |
Closest papers by meaning
- Human Motion Prediction via Learning Local Structure Representations and Temporal Dependencies (AAAI 2019 · similarity 0.90)
- Modeling Temporal Dynamics and Spatial Configurations of Actions Using Two-Stream Recurrent Neural Networks (CVPR 2017 · similarity 0.90)
- An End-to-End Spatio-Temporal Attention Model for Human Action Recognition from Skeleton Data (AAAI 2017 · similarity 0.90)
- Multi-level Motion Attention for Human Motion Prediction (IJCV 2021 · similarity 0.89)
- Learning Dynamic Relationships for 3D Human Motion Prediction (CVPR 2020 · similarity 0.89)
- Global-Local Motion Transformer for Unsupervised Skeleton-Based Action Learning (ECCV 2022 · similarity 0.89)
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.
