|
--- |
|
library_name: pytorch |
|
license: apache-2.0 |
|
tags: |
|
- android |
|
pipeline_tag: keypoint-detection |
|
|
|
--- |
|
|
|
 |
|
|
|
# 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:** Pose estimation |
|
- **Model Stats:** |
|
- Input resolution: 256x192 |
|
- Number of parameters: 17.9M |
|
- Model size: 68.5 MB |
|
|
|
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model |
|
|---|---|---|---|---|---|---|---|---| |
|
| RTMPose_Body2d | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 1.904 ms | 0 - 243 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 1.821 ms | 1 - 3 MB | FP16 | NPU | [RTMPose_Body2d.so](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.so) | |
|
| RTMPose_Body2d | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 2.444 ms | 0 - 109 MB | FP16 | NPU | [RTMPose_Body2d.onnx](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.onnx) | |
|
| RTMPose_Body2d | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 1.411 ms | 0 - 61 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 1.381 ms | 1 - 19 MB | FP16 | NPU | [RTMPose_Body2d.so](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.so) | |
|
| RTMPose_Body2d | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 1.725 ms | 0 - 24 MB | FP16 | NPU | [RTMPose_Body2d.onnx](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.onnx) | |
|
| RTMPose_Body2d | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 1.176 ms | 0 - 36 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 1.185 ms | 0 - 20 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 1.521 ms | 1 - 21 MB | FP16 | NPU | [RTMPose_Body2d.onnx](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.onnx) | |
|
| RTMPose_Body2d | SA7255P ADP | SA7255P | TFLITE | 46.39 ms | 1 - 29 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | SA7255P ADP | SA7255P | QNN | 46.336 ms | 1 - 8 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 1.909 ms | 0 - 242 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | SA8255 (Proxy) | SA8255P Proxy | QNN | 1.816 ms | 1 - 2 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | SA8295P ADP | SA8295P | TFLITE | 3.641 ms | 0 - 35 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | SA8295P ADP | SA8295P | QNN | 3.546 ms | 0 - 14 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 1.911 ms | 0 - 252 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | SA8650 (Proxy) | SA8650P Proxy | QNN | 1.811 ms | 1 - 3 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | SA8775P ADP | SA8775P | TFLITE | 3.188 ms | 0 - 30 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | SA8775P ADP | SA8775P | QNN | 3.009 ms | 1 - 11 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | QCS8275 (Proxy) | QCS8275 Proxy | TFLITE | 46.39 ms | 1 - 29 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | QCS8275 (Proxy) | QCS8275 Proxy | QNN | 46.336 ms | 1 - 8 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 1.905 ms | 0 - 240 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 1.823 ms | 1 - 3 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | QCS9075 (Proxy) | QCS9075 Proxy | TFLITE | 3.188 ms | 0 - 30 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | QCS9075 (Proxy) | QCS9075 Proxy | QNN | 3.009 ms | 1 - 11 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 3.051 ms | 0 - 55 MB | FP16 | NPU | [RTMPose_Body2d.tflite](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.tflite) | |
|
| RTMPose_Body2d | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 3.061 ms | 0 - 23 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 1.915 ms | 1 - 1 MB | FP16 | NPU | Use Export Script | |
|
| RTMPose_Body2d | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 2.344 ms | 37 - 37 MB | FP16 | NPU | [RTMPose_Body2d.onnx](https://huggingface.co/qualcomm/RTMPose_Body2d/blob/main/RTMPose_Body2d.onnx) | |
|
|
|
|
|
|
|
|
|
## Installation |
|
|
|
|
|
Install the package via pip: |
|
```bash |
|
pip install "qai-hub-models[rtmpose-body2d]" torch==2.4.1 -f https://download.openmmlab.com/mmcv/dist/cpu/torch2.4/index.html -f https://qaihub-public-python-wheels.s3.us-west-2.amazonaws.com/index.html |
|
``` |
|
|
|
|
|
## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device |
|
|
|
Sign-in to [Qualcomm® AI Hub](https://app.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://app.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 |
|
``` |
|
``` |
|
Profiling Results |
|
------------------------------------------------------------ |
|
RTMPose_Body2d |
|
Device : Samsung Galaxy S23 (13) |
|
Runtime : TFLITE |
|
Estimated inference time (ms) : 1.9 |
|
Estimated peak memory usage (MB): [0, 243] |
|
Total # Ops : 256 |
|
Compute Unit(s) : NPU (256 ops) |
|
``` |
|
|
|
|
|
## 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 S24") |
|
|
|
# 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. [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 --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 -- --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). |
|
* The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf) |
|
|
|
|
|
|
|
## 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). |
|
|
|
|
|
|
