---
title: cv_vit-base_human-pose-estimation_cvpr23-cvf
canonical_url: "https://www.modelscope.cn/models/iic/cv_vit-base_human-pose-estimation_cvpr23-cvf"
md_url: "https://www.modelscope.cn/models/iic/cv_vit-base_human-pose-estimation_cvpr23-cvf.md"
repository: iic/cv_vit-base_human-pose-estimation_cvpr23-cvf
chinese_name: "3D人体姿态估计模型"
last_updated: 2023-08-27
library_name:
  - pytorch
frameworks:
  - pytorch
downloads: 250
stars: 3
---

# cv_vit-base_human-pose-estimation_cvpr23-cvf

> cv_vit-base_human-pose-estimation_cvpr23-cvf - iic 在 ModelScope 开源的模型。人体SMPL参数估计模型

- **Repository**: iic/cv_vit-base_human-pose-estimation_cvpr23-cvf
- **Downloads**: 250
- **Stars**: 3
- **Last updated**: 2023-08-27

Source: https://www.modelscope.cn/models/iic/cv_vit-base_human-pose-estimation_cvpr23-cvf

---

输入一张人体图像，输出人体参数化模型SMPL的pose和shape估计结果，已经对应的vertices，3D关键点。
## SMPL模型
pose     维度(24, 3)，包含1个global rotation与23个local joints rotation，用三维的旋转向量表示

shape    维度(1, 10)

vertices 维度(6890, 3)，人体Mesh的6890个顶点，faces为SMPL模型的标准faces。
## 模型描述
模型来源CVPR 2023工作 Towards Stable Human Pose Estimation via Cross-View Fusion and Foot
Stabilization，由三部分组成：

1. ViT作为backbone用于提取图像特征
2. Cross-View Fusion module用于预测front-view 2D heatmaps，同时基于front-view 2D heatmaps与图像特征进一步得到side-view与top-view的2D heatmaps。
3. RKTD用于辅助脚部姿态预测。（本模型中不包含此模块）

![image.png](https://intranetproxy.alipay.com/skylark/lark/0/2023/png/301951/1691133647010-65a3af2e-fc2d-4ee2-a672-182b2d37a7c9.png#clientId=ua4640cae-2b2e-4&from=paste&height=578&id=u76ed4461&originHeight=795&originWidth=1541&originalType=binary&ratio=1.375&rotation=0&showTitle=false&size=804554&status=done&style=none&taskId=u826b67db-ff4c-49a1-b139-2683e58c725&title=&width=1120.7272727272727)
## 推理代码范例
## 引用
```python
@InProceedings{Zhuo_2023_CVPR,
    author    = {Zhuo, Li{\textquoteright}an and Cao, Jian and Wang, Qi and Zhang, Bang and Bo, Liefeng},
    title     = {Towards Stable Human Pose Estimation via Cross-View Fusion and Foot Stabilization},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2023},
    pages     = {650-659}
}
```