Research map / 3D vision / Human pose estimation
Capturing the Motion of Every Joint: 3D Human Pose and Shape Estimation with Independent Tokens
Read the paperCheck 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 INT-2 instead of MAED | 3DPW · 3DPW | PA-MPJPE: 45.7 -> 42.0 | Helped |
| Use INT-2 instead of MAED | Human3.6M · Human3.6M | PA-MPJPE: 38.7 -> 38.4 | No clear effect |
| Add temporal modeling to the image-based INT model (ablation) | INT on 3DPW · 3DPW | PA-MPJPE: 45.6 -> 43.9 | Helped |
| Add temporal modeling to the image-based INT model (ablation) | INT on 3DPW · 3DPW | Accel: 23.5 -> 19.1 | Helped |
| Use separate joint temporal learning rather than whole-pose temporal learning (ablation) | INT on 3DPW · 3DPW | PA-MPJPE: 43.9 -> 42.0 | Helped |
| Use separate joint temporal learning rather than whole-pose temporal learning (ablation) | INT on 3DPW · 3DPW | Accel: 19.1 -> 16.5 | Helped |
| Use ResNet50+Transformer rather than ViT-base as the image model (ablation) | INT on 3DPW · 3DPW | PA-MPJPE: 51.6 -> 42.0 | Helped |
| Use separate joint temporal learning rather than parallel spatial-temporal modeling (ablation) | INT on 3DPW · 3DPW | PA-MPJPE: 43.0 -> 42.0 | Helped |
| Use separate joint temporal learning rather than parallel spatial-temporal modeling (ablation) | INT on 3DPW · 3DPW | Accel: 19.4 -> 16.5 | Helped |
| Increase the temporal input sequence length (ablation) | INT on 3DPW, comparing T=16 and T=128 · 3DPW | Accel: 18.1 -> 15.1 | Helped |
| Increase the temporal input sequence length from T=16 to T=128 (ablation) | INT on 3DPW · 3DPW | PA-MPJPE: 42.2 -> 42.2 | No clear effect |
| Train the model with 3DPW training data | INT · 3DPW | PA-MPJPE: 49.7 -> 42.0 | Helped |
Closest papers by meaning
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.92)
- Global-to-Local Modeling for Video-Based 3D Human Pose and Shape Estimation (CVPR 2023 · similarity 0.91)
- PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · similarity 0.91)
- 3D Human Pose Estimation Using Spatio-Temporal Networks with Explicit Occlusion Training (AAAI 2020 · similarity 0.90)
- TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representation (CVPR 2024 · similarity 0.89)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · 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.