Research map / 3D vision / Human pose estimation
Sapiens: Foundation for Human Vision Models
Rawal Khirodkar, Timur Bagautdinov, Julieta Martinez, Zhaoen Su, Austin T James, Peter Selednik, Stuart Anderson, Shunsuke Saito
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 |
|---|---|---|---|
| pretrain on human images (Humans-300M) instead of same-size general images (ablation) | Sapiens-0.3B · Humans-5K pose | mAP: 37.3 -> 47.0 | Improved |
| pretrain on human images instead of general images (ablation) | Sapiens-0.3B · Humans-2K segmentation | mIoU: 52.8 -> 66.5 | Improved |
| pretrain on human images instead of general images (ablation) | Sapiens-0.3B · THuman2.0 depth | RMSE: 0.347 -> 0.288 | Improved |
| pretrain on human images instead of general images (ablation) | Sapiens-0.3B · Hi4D normal | Mean angular error (deg): 26.8 -> 21.8 | Improved |
| scale human pretraining data from 100M to 300M images (ablation) | Sapiens-0.3B · Humans-5K pose | mAP: 43.6 -> 47.0 | Improved |
| scale general pretraining data from 100M to 300M images (ablation) | Sapiens-0.3B · Humans-5K pose | mAP: 35.7 -> 37.3 | Improved |
| add pretraining versus random initialization (ablation) | Sapiens-0.3B · Humans-2K segmentation | mIoU: 40.3 -> 66.5 | Improved |
| scale model from 0.3B to 2B parameters (ablation) | Sapiens · Humans-5K pose | Whole-body AP: 53.4 -> 61.1 | Improved |
| scale model from 0.3B to 2B parameters (ablation) | Sapiens · Humans-2K segmentation | mIoU: 76.7 -> 81.2 | Improved |
| scale model from 0.3B to 2B parameters (ablation) | Sapiens · Hi4D normal | Median angular error (deg): 9.59 (1B) vs 9.62 (2B) | No clear effect |
| use Sapiens-2B depth versus DepthAnything on Hi4D | Sapiens-2B · Hi4D depth | AbsRel: 0.036 vs 0.034 (DepthAny-B) | Mixed |
Later papers that compared with this method
5 later accepted papers report Sapiens 2B as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Skullptor: High Fidelity 3D Head Reconstruction in Seconds with Multi-View Normal Prediction (CVPR 2026 · better on 7 of 8 shared results)
- Face Anything: 4D Face Reconstruction from Any Image Sequence (ECCV 2026 · better on 4 of 4 shared results)
- Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction (ICLR 2026 · better on 1 of 1 shared result)
- DAViD: Data-efficient and Accurate Vision Models from Synthetic Data (ICCV 2025 · better on 5 of 8 shared results)
- Detection, Pose Estimation and Segmentation for Multiple Bodies: Closing the Virtuous Circle (ICCV 2025 · better on 2 of 2 shared results)
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.
- ZoomNAS — ZoomNAS: Searching for Whole-body Human Pose Estimation in the Wild (TPAMI 2022) · 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
- SapiensID: Foundation for Human Recognition (CVPR 2025 · similarity 0.92)
- Evaluating Multiview Object Consistency in Humans and Image Models (NeurIPS 2024 · similarity 0.89)
- HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion (ICLR 2024 · similarity 0.88)
- SapiensID 2.0: Aligning Human Recognition Foundation Models with Human Perception (NeurIPS 2026 · similarity 0.87)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.87)
- Not Too Generative, Not Too Discriminative: The Human Alignment Sweet Spot (NeurIPS 2026 · 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.
