Research map / 3D vision / Human pose estimation
Diffusion Model is a Good Pose Estimator from 3D RF-Vision
Junqiao Fan, Jianfei Yang, Yuecong Xu, Lihua Xie
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 radar-conditioned diffusion refinement (without radar guidance) (ablation) | P4Transformer coarse poses · mmBody average | MPJPE (mm): 78.10 -> 73.31 | Improved |
| add global radar context with joint-wise features (ablation) | diffusion model · mmBody average | MPJPE (mm): 73.31 -> 70.99 | Improved |
| use joint feature template instead of direct PC features for global guidance (ablation) | GRC · mmBody average | MPJPE (mm): 72.50 -> 70.99 | Improved |
| use dynamic joint anchors instead of static anchors in local radar context (ablation) | LRC · mmBody average | MPJPE (mm): 71.05 -> 70.43 | Improved |
| add temporal motion consistency (TMC) (ablation) | mmDiff(G,L) · mmBody average | MPJPE (mm): 69.79 -> 69.16 | Improved |
| add structural limb-length consistency (SLC) (ablation) | mmDiff(G,L,T) · mmBody average | MPJPE (mm): 69.16 -> 68.08 | Improved |
| add limb-length module in Lab2 scene (ablation) | mmDiff(G,L,T) · mmBody Lab2 | MPJPE (mm): 68.12 -> 69.36 | Hurt |
| apply mmDiff to PointTransformer baseline on mm-Fi | PointTransformer · mm-Fi random split | MPJPE (mm): 73.09 -> 65.26 | Improved |
| apply mmDiff vs P4Transformer in adverse environments | P4Transformer · mmBody adverse scenes | MPJPE: 14.7% (MPJPE) and 12.0% (PA-MPJPE) improvement | Improved |
| diffusion modules vs benchmark network latency | mmDiff · mmBody efficiency | latency (ms): 40.48 ms -> 11.85 ms | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DiffPose — DiffPose: Toward More Reliable 3D Pose Estimation (CVPR 2023) · better on 2 of 2 shared results
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
- MVDoppler-Pose: Multi-Modal Multi-View mmWave Sensing for Long-Distance Self-Occluded Human Walking Pose Estimation (CVPR 2025 · similarity 0.89)
- RAPTR: Radar-based 3D Pose Estimation using Transformer (NeurIPS 2025 · similarity 0.85)
- $\text{Di}^2\text{Pose}$: Discrete Diffusion Model for Occluded 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.84)
- MonoDiff: Monocular 3D Object Detection and Pose Estimation with Diffusion Models (CVPR 2024 · similarity 0.84)
- Object Pose Estimation via the Aggregation of Diffusion Features (CVPR 2024 · similarity 0.83)
- Towards Reliable Identification of Diffusion-based Image Manipulations (NeurIPS 2025 · similarity 0.83)
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.
