Skip to content
💃 Real-time single person pose estimation for Android and iOS.
C++ Java Jupyter Notebook C Python Kotlin Other
Branch: master
Clone or download
Latest commit e31fb85 May 31, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
android_demo add gpu for tflite May 26, 2019
images update img Mar 28, 2019
release [+] ymls and new apks Jul 30, 2018
training update annotation Mar 28, 2019
.gitignore fix get_train_dataset_pipline to get_train_dataset_pipeline(I think i… Jul 25, 2018
LICENSE Update LICENSE May 12, 2018 add benchmark for iOS Apr 3, 2019

This repository currently implemented the CPM and Hourglass model using TensorFlow. Instead of normal convolution, inverted residuals (also known as Mobilenet V2) module has been used inside the model for real-time inference.

Model FLOPs PCKh Inference Time
CPM 0.5G 93.78 ~60 FPS on Snapdragon 845
~60 FPS on iPhone XS (need more test)
Hourglass 0.5G 91.81

You can modify the architectures of network for training much higher PCKh model.

Note: The repository only provide the baseline for mobile inference. Both model architectures (accuracy) and dataset still have a huge margin of improvement.

The respository contains:

  • Code of training cpm & hourglass model
  • Android demo source code (thanks to littleGnAl)
  • iOS demo source code (thanks to tucan)

Below GIF is the performance of Android and iOS

Android Mi Mix2s (~60 FPS) iPhone X (~30 FPS)
image image

You can download the apk as below to test on your device.

Using Mace (Support GPU) Using TFlite (Only CPU)
PoseEstimation-Mace.apk PoseEstimation-TFlite.apk

Issue and PR are welcome when you come across any problem.



  • Python3
  • TensorFlow >= 1.4
  • Mace


Training dataset available through google driver.

Unzip it will obtain the following file structure

# root @ ubuntu in ~/hdd/ai_challenger
$ tree -L 1 .
├── ai_challenger_train.json
├── ai_challenger_valid.json
├── train
└── valid

The traing dataset only contains single person images and it come from the competition of AI Challenger.

  • 22446 training examples
  • 1500 testing examples

I transfer the annotation into COCO format for using the data augument code from tf-pose-estimation respository.


In training procedure, we use cfg file on experiments folder for passing the hyper-parameter.

Below is the content of mv2_cpm.cfg.

model: 'mv2_cpm'
checkpoint: False
datapath: '/root/hdd/ai_challenger'
imgpath: '/root/hdd/'
visible_devices: '0, 1, 2'
multiprocessing_num: 8
max_epoch: 1000
lr: '0.001'
batchsize: 5
decay_rate: 0.95
input_width: 192
input_height: 192
n_kpoints: 14
scale: 2
modelpath: '/root/hdd/trained/mv2_cpm/models'
logpath: '/root/hdd/trained/mv2_cpm/log'
num_train_samples: 20000
per_update_tensorboard_step: 500
per_saved_model_step: 2000
pred_image_on_tensorboard: True

The cfg not cover all the parameters of the model, there still have some parameters in the

Train by nvidia-docker

Build the docker by the following command:

cd training/docker
docker build -t single-pose .


docker pull edvardhua/single-pose

Then run the following command to train the model:

nvidia-docker run -it -d \
-v <dataset_path>:/data5 -v <training_code_path>/training:/workspace \
-p 6006:6006 -e LOG_PATH=/root/hdd/trained/mv2_cpm/log \
-e PARAMETERS_FILE=experiments/mv2_cpm.cfg edvardhua/single-pose

Also, it will create the tensorboard on port 6006. Beside, make sure you install the nvidia-docker.

Train by ordinary way

  1. (mac only) Change tensorflow-gpu==1.4.0 to tensorflow==1.4.0 on requirements.txt.

  2. install the dependencies.

cd training
pip3 install -r requirements.txt

Beside, you also need to install cocoapi

  1. Edit the parameters files in experiments folder, it contains almost all the hyper-parameters and other configuration you need to define in training. After that, passing the parameters file to start the training:
cd training
python3 src/ experiments/mv2_cpm.cfg

After 12 hour training, the model is almost coverage on 3 Nvidia 1080Ti graphics cards, below is the corresponding plot on tensorboard.


Bechmark (PCKh)

Run the follow command to evaluate the value of your PCKh.

python3 src/ --frozen_pb_path=hourglass/model-360000.pb \
--anno_json_path=/root/hdd/ai_challenger/ai_challenger_valid.json \
--img_path=/root/hdd \

Pretain model



Android Demo

Thanks to mace framework, now you can using GPU to run this model on android smartphone.

Following command can transfer model into mace format.

cd <your-mace-path>
# You transer hourglass or cpm model by changing `yml` file.
python tools/ convert --config=<PoseEstimationForMobilePath>/release/mace_ymls/cpm.yml

Then follow the instruction of mace-0.9 documentation to integrate into android.

For how to invoke the model and parsing output, you can check the android source code i provided.

The benchmark of some socs for average inference time are shown as follow.

Model Snapdragon 845 Snapdragon 660 Hisilicon 960 Exynos 7420
CPM & Hourglass 17 ms 30 ms 42 ms 103 ms

Below is the environments i build this demo.

  • Operation System: macOS 10.13.6 (mace not support build under windows now)
  • Android Studio: 3.0.1
  • NDK Version: r16
  • Mace Version: 0.9.0

Different environments may encounter different error when you build mace-demo. To avoid this, i suggest using docker.

docker pull

docker run -it
	--privileged -d --name mace-dev 
	-v to/you/path/PoseEstimationForMobile/android_demo/demo_mace:/demo_mace

docker run -it --privileged -d --name mace-dev --net=host \
           -v to/you/path/PoseEstimationForMobile/android_demo/demo_mace:/demo_mace  \

# Enter to docker
docker exec -it mace-dev bash

# Exec command inside the docker
cd /demo_mace && ./gradlew build

Or you can transfer the model into tflite.

# Convert to frozen pb.
cd training
python3 src/ \
--checkpoint=<you_training_model_path>/model-xxx --output_graph=<you_output_model_path>/model-xxx.pb \
--size=192 --model=mv2_cpm_2

# If you update tensorflow to 1.9, run following command.
python3 src/ \
--frozen_pb=forzen_graph.pb \
--input_node_name='image' \
--output_node_name='Convolutional_Pose_Machine/stage_5_out' \
--output_path='./' \
# Convert to tflite.
# See for more information.
bazel-bin/tensorflow/contrib/lite/toco/toco \
--input_file=<you_output_model_path>/model-xxx.pb \
--output_file=<you_output_tflite_model_path>/mv2-cpm.tflite \
--input_format=TENSORFLOW_GRAPHDEF --output_format=TFLITE \
--inference_type=FLOAT \
--input_shape="1,192,192,3" \
--input_array='image' \

Then, place the tflite file in android_demo/app/src/main/assets and modify the parameters in ImageClassifierFloatInception.kt.

    // parameters need to modify in ImageClassifierFloatInception.kt
     * Create ImageClassifierFloatInception instance
     * @param imageSizeX Get the image size along the x axis.
     * @param imageSizeY Get the image size along the y axis.
     * @param outputW The output width of model
     * @param outputH The output height of model
     * @param modelPath Get the name of the model file stored in Assets.
     * @param numBytesPerChannel Get the number of bytes that is used to store a single
     * color channel value.
    fun create(
      activity: Activity,
      imageSizeX: Int = 192,
      imageSizeY: Int = 192,
      outputW: Int = 96,
      outputH: Int = 96,
      modelPath: String = "mv2-cpm.tflite",
      numBytesPerChannel: Int = 4
    ): ImageClassifierFloatInception =

Finally, import the project to Android Studio and run in you smartphone.

iOS Demo

Thanks to tucan, now you can run model on iOS.

First, convert model into CoreML model.

# Convert to frozen pb.
cd training
python3 src/ \
--checkpoint=<you_training_model_path>/model-xxx --output_graph=<you_output_model_path>/model-xxx.pb \
--size=192 --model=mv2_cpm_2

# Run the following command to get mlmodel
python3 src/ \
--frozen_pb=forzen_graph.pb \
--input_node_name='image' \
--output_node_name='Convolutional_Pose_Machine/stage_5_out' \
--output_path='./' \

Then, follow the instruction on PoseEstimation-CoreML.

The benchmark of some socs for average inference time are shown as follow.

Model iPhone XS Max iPhone XS iPhone X iPhone 8 Plus iPhone 8
CPM & Hourglass 17 ms 16 ms 69 ms 64 ms 42 ms
Model iPhone 7 iPad Pro
iPhone SE iPad
iPhone 6 Plus
CPM & Hourglass 74 ms 41 ms 103 ms 118 ms 331 ms


[1] Paper of Convolutional Pose Machines
[2] Paper of Stack Hourglass
[3] Paper of MobileNet V2
[4] Repository PoseEstimation-CoreML
[5] Repository of tf-pose-estimation
[6] Devlope guide of TensorFlow Lite
[7] Mace documentation


Apache License 2.0

You can’t perform that action at this time.