README.md

---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: keypoint-detection

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/rtmpose_body2d/web-assets/model_demo.png)

# RTMPose-Body2d: Optimized for Mobile Deployment
## Human pose estimation


RTMPose is a machine learning model that detects human pose and returns a location and confidence for each of 133 joints.

This model is an implementation of RTMPose-Body2d found [here](https://github.com/open-mmlab/mmpose/tree/main/projects/rtmpose).


This repository provides scripts to run RTMPose-Body2d on Qualcomm® devices.
More details on model performance across various devices, can be found
[here](https://aihub.qualcomm.com/models/rtmpose_body2d).



### Model Details

- **Model Type:** Model_use_case.pose_estimation
- **Model Stats:**
  - Input resolution: 256x192
  - Number of parameters: 17.9M
  - Model size (float): 68.5 MB
  - Model size (w8a16): 18.2 MB

| Model | Precision | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit | Target Model
|---|---|---|---|---|---|---|---|---|
| RTMPose-Body2d | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | TFLITE | 7.523 ms | 0 - 139 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 7.508 ms | 1 - 124 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | TFLITE | 3.542 ms | 0 - 189 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | QCS8450 (Proxy) | Qualcomm® QCS8450 (Proxy) | QNN_DLC | 3.535 ms | 1 - 160 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | TFLITE | 1.714 ms | 0 - 3 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.792 ms | 1 - 2 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 2.272 ms | 0 - 41 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.onnx.zip) |
| RTMPose-Body2d | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | TFLITE | 2.491 ms | 0 - 138 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 2.505 ms | 1 - 125 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | SA7255P ADP | Qualcomm® SA7255P | TFLITE | 7.523 ms | 0 - 139 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 7.508 ms | 1 - 124 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | SA8295P ADP | Qualcomm® SA8295P | TFLITE | 3.551 ms | 0 - 148 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | SA8295P ADP | Qualcomm® SA8295P | QNN_DLC | 3.577 ms | 0 - 133 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | SA8775P ADP | Qualcomm® SA8775P | TFLITE | 2.491 ms | 0 - 138 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 2.505 ms | 1 - 125 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | TFLITE | 1.303 ms | 0 - 182 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 1.327 ms | 1 - 151 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.714 ms | 0 - 125 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.onnx.zip) |
| RTMPose-Body2d | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | TFLITE | 1.099 ms | 0 - 140 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 1.085 ms | 0 - 128 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 1.381 ms | 0 - 100 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.onnx.zip) |
| RTMPose-Body2d | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | TFLITE | 0.911 ms | 0 - 141 MB | NPU | [RTMPose-Body2d.tflite](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.tflite) |
| RTMPose-Body2d | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 0.915 ms | 1 - 128 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 1.262 ms | 1 - 101 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.onnx.zip) |
| RTMPose-Body2d | float | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.909 ms | 1 - 1 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.dlc) |
| RTMPose-Body2d | float | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2.268 ms | 36 - 36 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d.onnx.zip) |
| RTMPose-Body2d | w8a16 | Dragonwing Q-6690 MTP | Qualcomm® QCM6690 | ONNX | 89.299 ms | 45 - 60 MB | CPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | Dragonwing RB3 Gen 2 Vision Kit | Qualcomm® QCS6490 | ONNX | 176.198 ms | 43 - 55 MB | CPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | QCS8275 (Proxy) | Qualcomm® QCS8275 (Proxy) | QNN_DLC | 3.841 ms | 0 - 140 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | QNN_DLC | 1.735 ms | 0 - 2 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | QCS8550 (Proxy) | Qualcomm® QCS8550 (Proxy) | ONNX | 2.478 ms | 0 - 22 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | QCS9075 (Proxy) | Qualcomm® QCS9075 (Proxy) | QNN_DLC | 2.109 ms | 0 - 140 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | SA7255P ADP | Qualcomm® SA7255P | QNN_DLC | 3.841 ms | 0 - 140 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | SA8775P ADP | Qualcomm® SA8775P | QNN_DLC | 2.109 ms | 0 - 140 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | QNN_DLC | 1.221 ms | 0 - 174 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 Mobile | ONNX | 1.711 ms | 0 - 153 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | QNN_DLC | 0.886 ms | 0 - 143 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | Samsung Galaxy S25 | Snapdragon® 8 Elite For Galaxy Mobile | ONNX | 1.355 ms | 0 - 122 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | Snapdragon 7 Gen 4 QRD | Snapdragon® 7 Gen 4 Mobile | ONNX | 84.732 ms | 47 - 64 MB | CPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | QNN_DLC | 0.75 ms | 0 - 142 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | Snapdragon 8 Elite Gen 5 QRD | Snapdragon® 8 Elite Gen 5 Mobile | ONNX | 1.261 ms | 0 - 122 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |
| RTMPose-Body2d | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN_DLC | 1.951 ms | 0 - 0 MB | NPU | [RTMPose-Body2d.dlc](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.dlc) |
| RTMPose-Body2d | w8a16 | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2.663 ms | 19 - 19 MB | NPU | [RTMPose-Body2d.onnx.zip](https://huggingface.co/qualcomm/RTMPose-Body2d/blob/main/RTMPose-Body2d_w8a16.onnx.zip) |




## Installation


Install the package via pip:
```bash
# NOTE: 3.10 <= PYTHON_VERSION < 3.14 is supported.
pip install mmpose==1.2.0 --no-deps
pip install "qai-hub-models[rtmpose-body2d]"
```


## Configure Qualcomm® AI Hub Workbench to run this model on a cloud-hosted device

Sign-in to [Qualcomm® AI Hub Workbench](https://workbench.aihub.qualcomm.com/) with your
Qualcomm® ID. Once signed in navigate to `Account -> Settings -> API Token`.

With this API token, you can configure your client to run models on the cloud
hosted devices.
```bash
qai-hub configure --api_token API_TOKEN
```
Navigate to [docs](https://workbench.aihub.qualcomm.com/docs/) for more information.



## Demo off target

The package contains a simple end-to-end demo that downloads pre-trained
weights and runs this model on a sample input.

```bash
python -m qai_hub_models.models.rtmpose_body2d.demo
```

The above demo runs a reference implementation of pre-processing, model
inference, and post processing.

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.rtmpose_body2d.demo
```


### Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm®
device. This script does the following:
* Performance check on-device on a cloud-hosted device
* Downloads compiled assets that can be deployed on-device for Android.
* Accuracy check between PyTorch and on-device outputs.

```bash
python -m qai_hub_models.models.rtmpose_body2d.export
```



## How does this work?

This [export script](https://aihub.qualcomm.com/models/rtmpose_body2d/qai_hub_models/models/RTMPose-Body2d/export.py)
leverages [Qualcomm® AI Hub](https://aihub.qualcomm.com/) to optimize, validate, and deploy this model
on-device. Lets go through each step below in detail:

Step 1: **Compile model for on-device deployment**

To compile a PyTorch model for on-device deployment, we first trace the model
in memory using the `jit.trace` and then call the `submit_compile_job` API.

```python
import torch

import qai_hub as hub
from qai_hub_models.models.rtmpose_body2d import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S25")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

```


Step 2: **Performance profiling on cloud-hosted device**

After compiling models from step 1. Models can be profiled model on-device using the
`target_model`. Note that this scripts runs the model on a device automatically
provisioned in the cloud.  Once the job is submitted, you can navigate to a
provided job URL to view a variety of on-device performance metrics.
```python
profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)
        
```

Step 3: **Verify on-device accuracy**

To verify the accuracy of the model on-device, you can run on-device inference
on sample input data on the same cloud hosted device.
```python
input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

```
With the output of the model, you can compute like PSNR, relative errors or
spot check the output with expected output.

**Note**: This on-device profiling and inference requires access to Qualcomm®
AI Hub Workbench. [Sign up for access](https://myaccount.qualcomm.com/signup).



## Run demo on a cloud-hosted device

You can also run the demo on-device.

```bash
python -m qai_hub_models.models.rtmpose_body2d.demo --eval-mode on-device
```

**NOTE**: If you want running in a Jupyter Notebook or Google Colab like
environment, please add the following to your cell (instead of the above).
```
%run -m qai_hub_models.models.rtmpose_body2d.demo -- --eval-mode on-device
```


## Deploying compiled model to Android


The models can be deployed using multiple runtimes:
- TensorFlow Lite (`.tflite` export): [This
  tutorial](https://www.tensorflow.org/lite/android/quickstart) provides a
  guide to deploy the .tflite model in an Android application.


- QNN (`.so` export ): This [sample
  app](https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/sample_app.html)
provides instructions on how to use the `.so` shared library  in an Android application.


## View on Qualcomm® AI Hub
Get more details on RTMPose-Body2d's performance across various devices [here](https://aihub.qualcomm.com/models/rtmpose_body2d).
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)


## License
* The license for the original implementation of RTMPose-Body2d can be found
  [here](https://github.com/open-mmlab/mmpose/blob/main/LICENSE).



## References
* [RTMPose: Real-Time Multi-Person Pose Estimation based on MMPose](https://arxiv.org/abs/2303.07399)
* [Source Model Implementation](https://github.com/open-mmlab/mmpose/tree/main/projects/rtmpose)



## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).