README.md

---
license: other
license_name: sapiens2-license
license_link: https://github.com/facebookresearch/sapiens2/blob/main/LICENSE.md
pipeline_tag: keypoint-detection
library_name: sapiens
base_model: facebook/sapiens2-pretrain-0.4b
tags:
  - sapiens
  - sapiens2
  - human-centric
  - pose
---

# Sapiens2-0.4B-Pose

308-keypoint top-down pose estimation including detailed face (274 keypoints), hand, and foot keypoints. Predictions follow the [Sociopticon keypoint format](https://github.com/facebookresearch/sapiens2/blob/main/sapiens/pose/configs/_base_/keypoints308.py).

This repository contains the **0.4B Pose Estimation** checkpoint, finetuned from the [Sapiens2-0.4B pretrained backbone](https://huggingface.co/facebook/sapiens2-pretrain-0.4b).

Pose is top-down — it requires bounding boxes from a person detector. We use [RTMDet](https://github.com/open-mmlab/mmdetection/tree/main/configs/rtmdet).

- 📄 **Paper:** [arXiv:2604.21681](https://arxiv.org/pdf/2604.21681)
- 🌐 **Project Page:** [rawalkhirodkar.github.io/sapiens2](https://rawalkhirodkar.github.io/sapiens2)
- 💻 **Code:** [github.com/facebookresearch/sapiens2](https://github.com/facebookresearch/sapiens2)

## Model Details

- **Developed by:** Meta
- **Model type:** Vision Transformer
- **License:** [Sapiens2 License](https://github.com/facebookresearch/sapiens2/blob/main/LICENSE.md)
- **Task:** pose
- **Base model:** [facebook/sapiens2-pretrain-0.4b](https://huggingface.co/facebook/sapiens2-pretrain-0.4b)
- **Format:** safetensors
- **File:** `sapiens2_0.4b_pose.safetensors`

## Quick Start

Install the [Sapiens2 repo](https://github.com/facebookresearch/sapiens2) (`pip install -e .`), download the checkpoint, and run the demo:

```bash
# 1. Download the checkpoint to $SAPIENS_CHECKPOINT_ROOT/pose/
hf download facebook/sapiens2-pose-0.4b sapiens2_0.4b_pose.safetensors \
    --local-dir ~/sapiens2_host/pose

# 2. Run the demo (edit INPUT, OUTPUT, and MODEL_NAME inside the script)
cd $SAPIENS_ROOT/sapiens/pose
./scripts/demo/keypoints308.sh
```

See the [Pose Estimation guide](https://github.com/facebookresearch/sapiens2/blob/main/docs/POSE.md) for details on inputs, outputs, and visualization options.

## Model Card

| Field | Value |
|-------|-------|
| Architecture | Sapiens2 ViT backbone + Pose Estimation head |
| Backbone parameters | 0.398 B |
| Backbone FLOPs | 1.260 T |
| Embedding dim | 1024 |
| Layers | 24 |
| Attention heads | 16 |
| Inference resolution | 1024 × 768 (H × W) |
| Patch size | 16 |

### Sapiens2-Pose Family

| Model | Params | FLOPs | Embed dim | Layers | Heads |
|-------|--------|-------|-----------|--------|-------|
| **Sapiens2-0.4B** *(this)* | 0.398 B | 1.260 T | 1024 | 24 | 16 |
| [Sapiens2-0.8B](https://huggingface.co/facebook/sapiens2-pose-0.8b) | 0.818 B | 2.592 T | 1280 | 32 | 16 |
| [Sapiens2-1B](https://huggingface.co/facebook/sapiens2-pose-1b) | 1.462 B | 4.715 T | 1536 | 40 | 24 |
| [Sapiens2-5B](https://huggingface.co/facebook/sapiens2-pose-5b) | 5.071 B | 15.722 T | 2432 | 56 | 32 |

See the [Sapiens2 Collection](https://huggingface.co/collections/facebook/sapiens2) for all variants and other downstream task checkpoints.

## Intended Use

- Pose Estimation on human-centric imagery
- Research on human-centric vision

## License

Released under the [Sapiens2 License](https://github.com/facebookresearch/sapiens2/blob/main/LICENSE.md).

## Citation

```bibtex
@article{khirodkarsapiens2,
  title={Sapiens2},
  author={Khirodkar, Rawal and Wen, He and Martinez, Julieta and Dong, Yuan and Su, Zhaoen and Saito, Shunsuke},
  journal={arXiv preprint arXiv:2604.21681},
  year={2026}
}
```