Research map / 3D vision / Human pose estimation
PoseBERT: A Generic Transformer Module for Temporal 3D Human Modeling
Fabien Baradel, Romain Brégier, Thibault Groueix, Philippe Weinzaepfel, Yannis Kalantidis, Grégory Rogez
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 PoseBERT on top of SPIN | SPIN · 3DPW | PA-MPJPE: 59.6 -> 57.3 | Improved |
| add PoseBERT on top of MoCap-SPIN | MoCap-SPIN · 3DPW | PA-MPJPE: 55.6 -> 52.9 | Improved |
| add PoseBERT on top of HybrIK with Faster-RCNN boxes | HybrIK · 3DPW | MPJPE: 98.1 -> 76.6 | Improved |
| add PoseBERT on top of HybrIK with ground-truth boxes | HybrIK · 3DPW | MPJPE: 75.2 -> 74.6 | Improved |
| add bi-GRU instead of PoseBERT with ground-truth boxes | HybrIK · 3DPW | MPJPE: 75.2 -> 83.9 | Hurt |
| add PoseBERT on top of LCR-Net++ without GT boxes | LCR-Net++ · MuPoTS-3D | MPJPE: 153.76 -> 126.62 | Improved |
| add PoseBERT instead of Savitzky-Golay filtering | LCR-Net++ · MuPoTS-3D | MPJPE: 138.54 -> 126.62 | Improved |
| add PoseBERT on top of HR-Net hand | HR-Net · DexYCB | MPJPE: 17.34 -> 14.05 | Improved |
| mask 12.5% of input in training (ablation) | PoseBERT · 3DPW | Accel: 9.6 -> 7.8 | Improved |
| add Gaussian noise to masking training (ablation) | PoseBERT · 3DPW | PA-MPJPE: 53.2 -> 52.9 | Improved |
| add positional encoding (ablation) | PoseBERT on MoCap-SPIN · 3DPW | PA-MPJPE: 54.8 -> 53.2 | Improved |
| unroll regressor across layers rather than one on top (ablation) | PoseBERT on MoCap-SPIN · 3DPW | PA-MPJPE: 54.5 -> 53.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- SPIN — SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition (AAAI 2021) · better on 2 of 2 shared results
- VIBE — VIBE: Video Inference for Human Body Pose and Shape Estimation (CVPR 2020) · better on 2 of 2 shared results
- LCR-Net++ — LCR-Net++: Multi-person 2D and 3D Pose Detection in Natural Images (TPAMI 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · similarity 0.90)
- MotionBERT: A Unified Perspective on Learning Human Motion Representations (ICCV 2023 · similarity 0.90)
- 3D Human Pose Estimation With Spatial and Temporal Transformers (ICCV 2021 · similarity 0.89)
- Capturing the Motion of Every Joint: 3D Human Pose and Shape Estimation with Independent Tokens (ICLR 2023 · similarity 0.89)
- Global-to-Local Modeling for Video-Based 3D Human Pose and Shape Estimation (CVPR 2023 · similarity 0.88)
- HiPoser: 3D Human Pose Estimation with Hierarchical Shared Learning at Parts-Level Using Inertial Measurement Units (AAAI 2025 · similarity 0.87)
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.
