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: 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 Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Precision Primary Compute Unit Target Model
PoseDetector Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 0.781 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 0.786 ms 0 - 18 MB FP16 NPU MediaPipe-Pose-Estimation.so
PoseDetector Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 0.917 ms 0 - 12 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseDetector Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 0.567 ms 0 - 25 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 0.57 ms 0 - 33 MB FP16 NPU MediaPipe-Pose-Estimation.so
PoseDetector Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 0.671 ms 0 - 36 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseDetector Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 0.46 ms 0 - 18 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 0.454 ms 0 - 21 MB FP16 NPU Use Export Script
PoseDetector Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 0.567 ms 0 - 19 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseDetector SA7255P ADP SA7255P TFLITE 37.869 ms 0 - 15 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector SA7255P ADP SA7255P QNN 37.832 ms 0 - 10 MB FP16 NPU Use Export Script
PoseDetector SA8255 (Proxy) SA8255P Proxy TFLITE 0.786 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector SA8255 (Proxy) SA8255P Proxy QNN 0.756 ms 0 - 3 MB FP16 NPU Use Export Script
PoseDetector SA8295P ADP SA8295P TFLITE 2.357 ms 0 - 19 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector SA8295P ADP SA8295P QNN 2.321 ms 0 - 18 MB FP16 NPU Use Export Script
PoseDetector SA8650 (Proxy) SA8650P Proxy TFLITE 0.786 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector SA8650 (Proxy) SA8650P Proxy QNN 0.757 ms 0 - 11 MB FP16 NPU Use Export Script
PoseDetector SA8775P ADP SA8775P TFLITE 1.79 ms 0 - 15 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector SA8775P ADP SA8775P QNN 1.734 ms 0 - 10 MB FP16 NPU Use Export Script
PoseDetector QCS8275 (Proxy) QCS8275 Proxy TFLITE 37.869 ms 0 - 15 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector QCS8275 (Proxy) QCS8275 Proxy QNN 37.832 ms 0 - 10 MB FP16 NPU Use Export Script
PoseDetector QCS8550 (Proxy) QCS8550 Proxy TFLITE 0.785 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector QCS8550 (Proxy) QCS8550 Proxy QNN 0.751 ms 0 - 3 MB FP16 NPU Use Export Script
PoseDetector QCS9075 (Proxy) QCS9075 Proxy TFLITE 1.79 ms 0 - 15 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector QCS9075 (Proxy) QCS9075 Proxy QNN 1.734 ms 0 - 10 MB FP16 NPU Use Export Script
PoseDetector QCS8450 (Proxy) QCS8450 Proxy TFLITE 1.973 ms 0 - 26 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseDetector QCS8450 (Proxy) QCS8450 Proxy QNN 1.976 ms 0 - 20 MB FP16 NPU Use Export Script
PoseDetector Snapdragon X Elite CRD Snapdragon® X Elite QNN 0.886 ms 0 - 0 MB FP16 NPU Use Export Script
PoseDetector Snapdragon X Elite CRD Snapdragon® X Elite ONNX 0.942 ms 3 - 3 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 0.824 ms 0 - 71 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 0.81 ms 1 - 36 MB FP16 NPU MediaPipe-Pose-Estimation.so
PoseLandmarkDetector Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 1.119 ms 0 - 33 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 0.614 ms 0 - 50 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 0.6 ms 1 - 45 MB FP16 NPU MediaPipe-Pose-Estimation.so
PoseLandmarkDetector Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 0.781 ms 0 - 53 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 0.52 ms 0 - 32 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 0.552 ms 0 - 24 MB FP16 NPU Use Export Script
PoseLandmarkDetector Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 0.681 ms 1 - 30 MB FP16 NPU MediaPipe-Pose-Estimation.onnx
PoseLandmarkDetector SA7255P ADP SA7255P TFLITE 17.165 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector SA7255P ADP SA7255P QNN 17.094 ms 1 - 10 MB FP16 NPU Use Export Script
PoseLandmarkDetector SA8255 (Proxy) SA8255P Proxy TFLITE 0.814 ms 0 - 68 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector SA8255 (Proxy) SA8255P Proxy QNN 0.8 ms 1 - 4 MB FP16 NPU Use Export Script
PoseLandmarkDetector SA8295P ADP SA8295P TFLITE 1.431 ms 0 - 25 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector SA8295P ADP SA8295P QNN 1.39 ms 0 - 18 MB FP16 NPU Use Export Script
PoseLandmarkDetector SA8650 (Proxy) SA8650P Proxy TFLITE 0.819 ms 0 - 71 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector SA8650 (Proxy) SA8650P Proxy QNN 0.817 ms 1 - 3 MB FP16 NPU Use Export Script
PoseLandmarkDetector SA8775P ADP SA8775P TFLITE 1.633 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector SA8775P ADP SA8775P QNN 1.606 ms 1 - 11 MB FP16 NPU Use Export Script
PoseLandmarkDetector QCS8275 (Proxy) QCS8275 Proxy TFLITE 17.165 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector QCS8275 (Proxy) QCS8275 Proxy QNN 17.094 ms 1 - 10 MB FP16 NPU Use Export Script
PoseLandmarkDetector QCS8550 (Proxy) QCS8550 Proxy TFLITE 0.815 ms 0 - 71 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector QCS8550 (Proxy) QCS8550 Proxy QNN 0.806 ms 1 - 4 MB FP16 NPU Use Export Script
PoseLandmarkDetector QCS9075 (Proxy) QCS9075 Proxy TFLITE 1.633 ms 0 - 27 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector QCS9075 (Proxy) QCS9075 Proxy QNN 1.606 ms 1 - 11 MB FP16 NPU Use Export Script
PoseLandmarkDetector QCS8450 (Proxy) QCS8450 Proxy TFLITE 1.194 ms 0 - 42 MB FP16 NPU MediaPipe-Pose-Estimation.tflite
PoseLandmarkDetector QCS8450 (Proxy) QCS8450 Proxy QNN 1.157 ms 1 - 37 MB FP16 NPU Use Export Script
PoseLandmarkDetector Snapdragon X Elite CRD Snapdragon® X Elite QNN 1.025 ms 1 - 1 MB FP16 NPU Use Export Script
PoseLandmarkDetector Snapdragon X Elite CRD Snapdragon® X Elite ONNX 1.152 ms 7 - 7 MB FP16 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                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 0.8                    
Estimated peak memory usage (MB): [0, 27]                
Total # Ops                     : 106                    
Compute Unit(s)                 : NPU (106 ops)          

------------------------------------------------------------
PoseLandmarkDetector
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 0.8                    
Estimated peak memory usage (MB): [0, 71]                
Total # Ops                     : 219                    
Compute Unit(s)                 : NPU (219 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

-

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