Research map / 3D vision / Human pose estimation
SMPLer-X: Scaling Up Expressive Human Pose and Shape Estimation
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 |
|---|---|---|---|
| increase training data from 5 to 32 datasets (ablation) | SMPLer-X-B · Five benchmarks | MPE: 103.5 -> 74.3 | Helped |
| scale backbone from ViT-S to ViT-L at 32 datasets (ablation) | SMPLer-X · Five benchmarks | MPE: 82.6 -> 66.2 | Helped |
| scale backbone from ViT-L to ViT-H at 32 datasets (ablation) | SMPLer-X · Five benchmarks | MPE: 66.2 -> 63.1 | Helped |
| train on top-10 ranked datasets instead of bottom-10 (ablation) | SMPLer-X-B · Five benchmarks | MPE: 115.10 -> 89.20 | Helped |
| train on top-5 ranked datasets instead of bottom-5 (ablation) | SMPLer-X-B · Five benchmarks | MPE: 115.61 -> 103.47 | Helped |
| finetune neck and head only instead of full network (ablation) | SMPLer-X-H32 · UBody | PVE (All): 51.3 -> 51.1 | No clear effect |
| finetune head only instead of neck+head (ablation) | SMPLer-X-H32 · UBody | PVE (All): 51.1 -> 54.2 | Hurt |
| add finetuning with selected datasets (ablation) | SMPLer-X-L20 · AGORA test | NMVE (All): 133.1 -> 107.2 | Helped |
| weighted sampling by dataset rank instead of balanced (ablation) | SMPLer-X-H32 · Five benchmarks | MPE: 63.08 -> 62.12 | No clear effect |
| concatenate datasets instead of balanced sampling (ablation) | SMPLer-X-H32 · Five benchmarks | MPE: 63.08 -> 63.32 | No clear effect |
| use generalist foundation model instead of single-dataset training | SMPLer-X-B20 · AGORA val, UBody, EgoBody, 3DPW | primary error | Helped |
Closest papers by meaning
- AiOS: All-in-One-Stage Expressive Human Pose and Shape Estimation (CVPR 2024 · similarity 0.88)
- Expressive Gaussian Human Avatars from Monocular RGB Video (NeurIPS 2024 · similarity 0.86)
- Reconstructing Close Human Interaction with Appearance and Proxemics Reasoning (CVPR 2025 · similarity 0.86)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.86)
- MUC: Mixture of Uncalibrated Cameras for Robust 3D Human Body Reconstruction (AAAI 2025 · similarity 0.86)
- MVHumanNet: A Large-scale Dataset of Multi-view Daily Dressing Human Captures (CVPR 2024 · similarity 0.86)
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.