Research map / 3D vision / Human pose estimation
Towards Balanced Multi-Modal Learning in 3D Human Pose Estimation
Mengshi Qi, Jiaxuan Peng, Xianlin Zhang, Huadong Ma
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 AWC loss to concatenation fusion | Concatenation fusion · MM-Fi Protocol 1 | MPJPE (mm): 53.87 -> 51.16 | Improved |
| add AWC loss to attention fusion | Attention fusion · MM-Fi Protocol 1 | MPJPE (mm): 53.35 -> 51.29 | Improved |
| add G-Blending to concatenation fusion | Concatenation fusion · MM-Fi Protocol 1 | MPJPE (mm): 53.87 -> 58.40 | Hurt |
| add OGM-GE to concatenation fusion | Concatenation fusion · MM-Fi Protocol 1 | MPJPE (mm): 53.87 -> 55.51 | Hurt |
| add AGM to concatenation fusion | Concatenation fusion · MM-Fi Protocol 1 | MPJPE (mm): 53.87 -> 55.80 | Hurt |
| add Modality-level resampling to concatenation fusion | Concatenation fusion · MM-Fi Protocol 1 | MPJPE (mm): 53.87 -> 53.24 | No clear effect |
| add mmWave and WiFi to RGB+LiDAR (ablation) | Concatenation fusion · MM-Fi Protocol 1 | MPJPE (mm): 52.93 -> 53.87 | Hurt |
| add regularization on inferior modalities (alphaI=10k) alongside superior (alphaS=10k) (ablation) | AWC loss · MM-Fi Protocol 1 | MPJPE (mm): 52.09 -> 51.88 | Improved |
| increase alphaS from 10k to 20k with alphaI=10k (ablation) | AWC loss · MM-Fi Protocol 1 | MPJPE (mm): 51.88 -> 51.16 | Improved |
| increase alphaI from 10k to 20k with alphaS=20k (ablation) | AWC loss · MM-Fi Protocol 1 | MPJPE (mm): 51.16 -> 51.69 | Hurt |
| use learning window K=20 epochs for AWC (ablation) | AWC loss · MM-Fi Protocol 1 | MPJPE (mm): 51.16 | Improved |
| add RGB to LiDAR+mmWave+WiFi style fusion: RGB alone vs all four modalities (ablation) | Concatenation fusion · MM-Fi Protocol 1 | PA-MPJPE (mm): 35.75 -> 35.09 | No clear effect |
Closest papers by meaning
- FreeMan: Towards Benchmarking 3D Human Pose Estimation under Real-World Conditions (CVPR 2024 · similarity 0.87)
- X-Fi: A Modality-Invariant Foundation Model for Multimodal Human Sensing (ICLR 2025 · similarity 0.86)
- Multi-View 3D Human Pose Estimation with Weakly Synchronized Images (AAAI 2025 · similarity 0.85)
- $\text{Di}^2\text{Pose}$: Discrete Diffusion Model for Occluded 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.85)
- Matrix3D: Large Photogrammetry Model All-in-One (CVPR 2025 · similarity 0.85)
- Lifting by Image – Leveraging Image Cues for Accurate 3D Human Pose Estimation (AAAI 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.
