Research map / 3D vision / Human pose estimation
Adaptive Multi-View and Temporal Fusing Transformer for 3D Human Pose Estimation
Hui Shuai, Lele Wu, Qingshan Liu
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 Confidence Attentive Aggregation instead of no confidence (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 29.4 -> 28.0 | Improved |
| concatenate 2D pose with confidence instead of ignoring it (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 29.4 -> 29.2 | Improved |
| use CAA instead of concatenating confidence (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 29.2 -> 28.0 | Improved |
| use MFT instead of conventional transformer for multi-view fusion (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 31.9 -> 28.0 | Improved |
| use MFT instead of point transformer (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 28.4 -> 28.0 | Improved |
| add transform matrix Tij to MFT (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 30.5 -> 28.0 | Improved |
| add MFT module (train on 2 views, test on 4 views) (ablation) | MTF-Transformer · Human3.6M generalization | MPJPE: 57.0 -> 39.0 | Improved |
| use random block mask rate 0.4 instead of 0 (ablation) | MTF-Transformer · Human3.6M, mean over 1-4 views | MPJPE: 89.0 -> 34.9 | Improved |
| use mask rate 1.0 instead of 0.4 (ablation) | MTF-Transformer · Human3.6M, mean over 1-4 views | MPJPE: 34.9 -> 107.2 | Hurt |
| add one synthesized view in training instead of none (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 28.4 -> 28.0 | Improved |
| add a second synthesized view (2 instead of 1) (ablation) | MTF-Transformer · Human3.6M, 4 views, T=1 | MPJPE: 28.0 -> 28.1 | Hurt |
| increase test sequence length from 1 to 27 frames (ablation) | MTF-Transformer · Human3.6M, 4 views | MPJPE: 27.46 -> 26.21 | Improved |
Closest papers by meaning
- SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · similarity 0.90)
- Multiple View Geometry Transformers for 3D Human Pose Estimation (CVPR 2024 · similarity 0.89)
- ExtPose: Robust and Coherent Pose Estimation by Extending ViTs (ICML 2025 · similarity 0.87)
- Lifting by Image – Leveraging Image Cues for Accurate 3D Human Pose Estimation (AAAI 2024 · similarity 0.86)
- FusionFormer: A Concise Unified Feature Fusion Transformer for 3D Pose Estimation (AAAI 2024 · similarity 0.86)
- Hourglass Tokenizer for Efficient Transformer-Based 3D Human Pose Estimation (CVPR 2024 · 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.
