Hugging Face's logo

qualcomm
/
RTMPose_Body2d

Keypoint Detection
PyTorch
LiteRT
ONNX
android
Model card Files Community

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.

This repository provides scripts to run RTMPose_Body2d on Qualcomm® devices. More details on model performance across various devices, can be found here.

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 (w8a8): 17.52MB
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 7.737 ms 0 - 30 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN 7.612 ms 1 - 10 MB NPU Use Export Script
RTMPose_Body2d float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 3.192 ms 0 - 60 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN 3.72 ms 0 - 28 MB NPU Use Export Script
RTMPose_Body2d float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 1.914 ms 0 - 230 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN 1.824 ms 1 - 3 MB NPU Use Export Script
RTMPose_Body2d float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 2.717 ms 0 - 32 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN 2.55 ms 1 - 15 MB NPU Use Export Script
RTMPose_Body2d float SA7255P ADP Qualcomm® SA7255P TFLITE 7.737 ms 0 - 30 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float SA7255P ADP Qualcomm® SA7255P QNN 7.612 ms 1 - 10 MB NPU Use Export Script
RTMPose_Body2d float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 1.918 ms 0 - 230 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN 1.827 ms 1 - 3 MB NPU Use Export Script
RTMPose_Body2d float SA8295P ADP Qualcomm® SA8295P TFLITE 3.683 ms 0 - 33 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float SA8295P ADP Qualcomm® SA8295P QNN 3.601 ms 0 - 18 MB NPU Use Export Script
RTMPose_Body2d float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 1.909 ms 0 - 227 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN 1.818 ms 1 - 3 MB NPU Use Export Script
RTMPose_Body2d float SA8775P ADP Qualcomm® SA8775P TFLITE 2.717 ms 0 - 32 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float SA8775P ADP Qualcomm® SA8775P QNN 2.55 ms 1 - 15 MB NPU Use Export Script
RTMPose_Body2d float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 1.926 ms 0 - 230 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN 1.82 ms 1 - 79 MB NPU Use Export Script
RTMPose_Body2d float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 1.999 ms 0 - 113 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 1.414 ms 0 - 61 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN 1.333 ms 0 - 25 MB NPU Use Export Script
RTMPose_Body2d float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 1.494 ms 0 - 30 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 1.339 ms 0 - 35 MB NPU RTMPose_Body2d.tflite
RTMPose_Body2d float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN 1.298 ms 0 - 20 MB NPU Use Export Script
RTMPose_Body2d float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 1.507 ms 1 - 23 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d float Snapdragon X Elite CRD Snapdragon® X Elite QNN 1.923 ms 1 - 1 MB NPU Use Export Script
RTMPose_Body2d float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.974 ms 37 - 37 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d w8a16 Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 2.851 ms 0 - 44 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d w8a16 Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 1.997 ms 0 - 50 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d w8a16 Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 2.056 ms 0 - 33 MB NPU RTMPose_Body2d.onnx
RTMPose_Body2d w8a16 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 3.065 ms 19 - 19 MB NPU RTMPose_Body2d.onnx

Installation

Install the package via pip:

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 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.

qai-hub configure --api_token API_TOKEN

Navigate to 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.

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.
python -m qai_hub_models.models.rtmpose_body2d.export

Profiling Results
------------------------------------------------------------
RTMPose_Body2d
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 7.7                                  
Estimated peak memory usage (MB): [0, 30]                              
Total # Ops                     : 256                                  
Compute Unit(s)                 : npu (256 ops) gpu (0 ops) cpu (0 ops)

How does this work?

This export script leverages Qualcomm® AI Hub 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.

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.

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.

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.

Run demo on a cloud-hosted device

You can also run the demo on-device.

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 provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app 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. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of RTMPose_Body2d can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Keypoint Detection
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support