Research map / 3D vision / Human pose estimation
Deformable Pose Traversal Convolution for 3D Action and Gesture Recognition
Junwu Weng, Mengyuan Liu, Xudong Jiang, Junsong Yuan
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 deformable offsets to pose traversal convolution (ablation) | Pose Traversal Conv · DHG Fine | accuracy (%): 77.1 -> 81.9 | Improved |
| add deformable offsets to pose traversal convolution (ablation) | Pose Traversal Conv · DHG Both-14 | accuracy (%): 81.1 -> 85.8 | Improved |
| add deformable offsets to pose traversal convolution (ablation) | Pose Traversal Conv · NTU-RGB+D cross-subject | accuracy (%): 76.1 -> 76.8 | Improved |
| add deformable offsets to pose traversal convolution (ablation) | Pose Traversal Conv · Berkeley MHAD | accuracy (%): 98.6 -> 100.0 | Improved |
| use pose traversal representation instead of single-chain flattening (ablation) | Pose Chain Conv · DHG Fine | accuracy (%): 76.2 -> 77.1 | Improved |
| use pose traversal representation instead of single-chain flattening (ablation) | Pose Chain Conv · DHG Both-14 | accuracy (%): 80.4 -> 81.1 | Mixed |
| use pose traversal representation instead of single-chain flattening (ablation) | Pose Chain Conv · Berkeley MHAD | accuracy (%): 96.4 -> 98.6 | Improved |
| add deformable offsets under severe joint noise (50% noisy joints) (ablation) | Pose Traversal Conv · Berkeley MHAD noise | accuracy: 7.64% | Improved |
| add deformable offsets under small noise (amplitude up to 5) (ablation) | Pose Traversal Conv · Berkeley MHAD small noise | accuracy (%): 96.36 -> 98.18 | Improved |
| use 3D instead of 2D pose input | D-Pose Traversal Conv · DHG Both-14 | accuracy (%): 85.1 -> 85.8 | No clear effect |
| use 3D instead of 2D pose input | Pose Traversal Conv · DHG Fine | accuracy (%): 74.6 -> 77.1 | Improved |
| add deformable offsets for the Pinch gesture (ablation) | Pose Traversal Conv · DHG-14 Pinch | accuracy: 49% -> 71.5% | Improved |
Closest papers by meaning
- Cross-Modal Learning with 3D Deformable Attention for Action Recognition (ICCV 2023 · similarity 0.86)
- Recurrent 3D Pose Sequence Machines (CVPR 2017 · similarity 0.85)
- Propagating LSTM: 3D Pose Estimation based on Joint Interdependency (ECCV 2018 · similarity 0.85)
- PA3D: Pose-Action 3D Machine for Video Recognition (CVPR 2019 · similarity 0.85)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.85)
- Exploiting temporal information for 3D human pose estimation (ECCV 2018 · 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.
