Hugging Face's logo

qualcomm
/
MediaPipe-Pose-Estimation

Keypoint Detection
PyTorch
LiteRT
ONNX
real_time
android
Model card Files Community
1

MediaPipe-Pose-Estimation: Optimized for Mobile Deployment

Detect and track human body poses in real-time images and video streams

The MediaPipe Pose Landmark Detector is a machine learning pipeline that predicts bounding boxes and pose skeletons of poses in an image.

This model is an implementation of MediaPipe-Pose-Estimation found here.

This repository provides scripts to run MediaPipe-Pose-Estimation 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: 256x256
    • Number of parameters (MediaPipePoseDetector): 815K
    • Model size (MediaPipePoseDetector): 3.14 MB
    • Number of parameters (MediaPipePoseLandmarkDetector): 3.37M
    • Model size (MediaPipePoseLandmarkDetector): 12.9 MB
Model Precision Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Primary Compute Unit Target Model
PoseDetector float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 5.19 ms 0 - 15 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN 5.152 ms 0 - 9 MB NPU Use Export Script
PoseDetector float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1.984 ms 0 - 23 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN 2.042 ms 0 - 23 MB NPU Use Export Script
PoseDetector float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.788 ms 0 - 27 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN 0.76 ms 0 - 3 MB NPU Use Export Script
PoseDetector float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1.502 ms 0 - 18 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN 1.431 ms 0 - 15 MB NPU Use Export Script
PoseDetector float SA7255P ADP Qualcomm® SA7255P TFLITE 5.19 ms 0 - 15 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float SA7255P ADP Qualcomm® SA7255P QNN 5.152 ms 0 - 9 MB NPU Use Export Script
PoseDetector float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.787 ms 0 - 27 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN 0.758 ms 0 - 2 MB NPU Use Export Script
PoseDetector float SA8295P ADP Qualcomm® SA8295P TFLITE 2.34 ms 0 - 20 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float SA8295P ADP Qualcomm® SA8295P QNN 2.306 ms 0 - 18 MB NPU Use Export Script
PoseDetector float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.783 ms 0 - 27 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN 0.761 ms 0 - 2 MB NPU Use Export Script
PoseDetector float SA8775P ADP Qualcomm® SA8775P TFLITE 1.502 ms 0 - 18 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float SA8775P ADP Qualcomm® SA8775P QNN 1.431 ms 0 - 15 MB NPU Use Export Script
PoseDetector float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 0.782 ms 0 - 28 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN 0.759 ms 0 - 16 MB NPU Use Export Script
PoseDetector float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 0.931 ms 0 - 13 MB NPU MediaPipe-Pose-Estimation.onnx
PoseDetector float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.559 ms 0 - 31 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN 0.558 ms 0 - 28 MB NPU Use Export Script
PoseDetector float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.648 ms 0 - 31 MB NPU MediaPipe-Pose-Estimation.onnx
PoseDetector float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 0.459 ms 0 - 23 MB NPU MediaPipe-Pose-Estimation.tflite
PoseDetector float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN 0.556 ms 0 - 19 MB NPU Use Export Script
PoseDetector float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 0.686 ms 0 - 19 MB NPU MediaPipe-Pose-Estimation.onnx
PoseDetector float Snapdragon X Elite CRD Snapdragon® X Elite QNN 0.897 ms 0 - 0 MB NPU Use Export Script
PoseDetector float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.951 ms 0 - 0 MB NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) TFLITE 17.2 ms 0 - 27 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float QCS8275 (Proxy) Qualcomm® QCS8275 (Proxy) QNN 17.095 ms 1 - 11 MB NPU Use Export Script
PoseLandmarkDetector float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) TFLITE 1.115 ms 0 - 48 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float QCS8450 (Proxy) Qualcomm® QCS8450 (Proxy) QNN 1.273 ms 1 - 43 MB NPU Use Export Script
PoseLandmarkDetector float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) TFLITE 0.811 ms 0 - 70 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float QCS8550 (Proxy) Qualcomm® QCS8550 (Proxy) QNN 0.831 ms 1 - 4 MB NPU Use Export Script
PoseLandmarkDetector float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) TFLITE 1.491 ms 0 - 29 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float QCS9075 (Proxy) Qualcomm® QCS9075 (Proxy) QNN 1.421 ms 1 - 16 MB NPU Use Export Script
PoseLandmarkDetector float SA7255P ADP Qualcomm® SA7255P TFLITE 17.2 ms 0 - 27 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float SA7255P ADP Qualcomm® SA7255P QNN 17.095 ms 1 - 11 MB NPU Use Export Script
PoseLandmarkDetector float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) TFLITE 0.817 ms 1 - 69 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float SA8255 (Proxy) Qualcomm® SA8255P (Proxy) QNN 0.817 ms 1 - 3 MB NPU Use Export Script
PoseLandmarkDetector float SA8295P ADP Qualcomm® SA8295P TFLITE 1.432 ms 0 - 26 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float SA8295P ADP Qualcomm® SA8295P QNN 1.386 ms 0 - 18 MB NPU Use Export Script
PoseLandmarkDetector float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) TFLITE 0.817 ms 0 - 69 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float SA8650 (Proxy) Qualcomm® SA8650P (Proxy) QNN 0.825 ms 1 - 3 MB NPU Use Export Script
PoseLandmarkDetector float SA8775P ADP Qualcomm® SA8775P TFLITE 1.491 ms 0 - 29 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float SA8775P ADP Qualcomm® SA8775P QNN 1.421 ms 1 - 16 MB NPU Use Export Script
PoseLandmarkDetector float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile TFLITE 0.819 ms 0 - 70 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile QNN 0.796 ms 1 - 35 MB NPU Use Export Script
PoseLandmarkDetector float Samsung Galaxy S23 Snapdragon® 8 Gen 2 Mobile ONNX 1.107 ms 0 - 28 MB NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile TFLITE 0.597 ms 0 - 50 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile QNN 0.591 ms 1 - 45 MB NPU Use Export Script
PoseLandmarkDetector float Samsung Galaxy S24 Snapdragon® 8 Gen 3 Mobile ONNX 0.771 ms 0 - 50 MB NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile TFLITE 0.554 ms 0 - 34 MB NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile QNN 0.471 ms 1 - 29 MB NPU Use Export Script
PoseLandmarkDetector float Snapdragon 8 Elite QRD Snapdragon® 8 Elite Mobile ONNX 0.765 ms 1 - 26 MB NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector float Snapdragon X Elite CRD Snapdragon® X Elite QNN 1.011 ms 1 - 1 MB NPU Use Export Script
PoseLandmarkDetector float Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.165 ms 7 - 7 MB NPU MediaPipe-Pose-Estimation.onnx

Installation

Install the package via pip:

pip install qai-hub-models

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.mediapipe_pose.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.mediapipe_pose.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.mediapipe_pose.export

Profiling Results
------------------------------------------------------------
PoseDetector
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 5.2                                  
Estimated peak memory usage (MB): [0, 15]                              
Total # Ops                     : 106                                  
Compute Unit(s)                 : npu (106 ops) gpu (0 ops) cpu (0 ops)

------------------------------------------------------------
PoseLandmarkDetector
Device                          : cs_8275 (ANDROID 14)                 
Runtime                         : TFLITE                               
Estimated inference time (ms)   : 17.2                                 
Estimated peak memory usage (MB): [0, 27]                              
Total # Ops                     : 219                                  
Compute Unit(s)                 : npu (219 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.mediapipe_pose 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.

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 MediaPipe-Pose-Estimation's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

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

References

Community

Downloads last month
599
Inference Providers NEW
Keypoint Detection
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support