/tf-pose-estimation

Deep Pose Estimation implemented using Tensorflow with Custom Architectures for fast inference.

Primary LanguagePureBasicApache License 2.0Apache-2.0

tf-pose-estimation

'Openpose' for human pose estimation have been implemented using Tensorflow. It also provides several variants that have made some changes to the network structure for real-time processing on the CPU or low-power embedded devices.

You can even run this on your macbook with descent FPS!

Original Repo(Caffe) : https://github.com/CMU-Perceptual-Computing-Lab/openpose

CMU's Original Model
on Macbook Pro 15"
Mobilenet Variant
on Macbook Pro 15"
Mobilenet Variant
on Jetson TX2
cmu-model mb-model-macbook mb-model-tx2
~0.6 FPS ~4.2 FPS @ 368x368 ~10 FPS @ 368x368
2.8GHz Quad-core i7 2.8GHz Quad-core i7 Jetson TX2 Embedded Board

Implemented features are listed here : features

Important Updates

2018.5.21 Post-processing part is implemented in c++. It is required compiling the part. See: https://github.com/ildoonet/tf-pose-estimation/tree/master/src/pafprocess 2018.2.7 Arguments in run.py script changed. Support dynamic input size.

Install

Dependencies

You need dependencies below.

  • python3
  • tensorflow 1.4.1+
  • opencv3, protobuf, python3-tk

Opensources

Install

Clone the repo and install 3rd-party libraries.

$ git clone https://www.github.com/ildoonet/tf-openpose
$ cd tf-openpose
$ pip3 install -r requirements.txt

Build c++ library for post processing. See : https://github.com/ildoonet/tf-pose-estimation/tree/master/tf_pose/pafprocess

$ cd tf_pose/pafprocess
$ swig -python -c++ pafprocess.i && python3 setup.py build_ext --inplace

Package Install

Alternatively, you can install this repo as a shared package using pip.

$ git clone https://www.github.com/ildoonet/tf-openpose
$ cd tf-openpose
$ python setup.py install

Test installed package

package_install_result

python -c 'import tf_pose; tf_pose.infer(image="./images/p1.jpg")'

Models

I have tried multiple variations of models to find optmized network architecture. Some of them are below and checkpoint files are provided for research purpose.

  • cmu

    • the model based VGG pretrained network which described in the original paper.
    • I converted Weights in Caffe format to use in tensorflow.
    • pretrained weight download
  • dsconv

    • Same architecture as the cmu version except for the depthwise separable convolution of mobilenet.
    • I trained it using 'transfer learning', but it provides not-enough speed and accuracy.
  • mobilenet

    • Based on the mobilenet paper, 12 convolutional layers are used as feature-extraction layers.
    • To improve on small person, minor modification on the architecture have been made.
    • Three models were learned according to network size parameters.
    • I published models which is not the best ones, but you can test them before you trained a model from the scratch.

Download Tensorflow Graph File(pb file)

Before running demo, you should download graph files. You can deploy this graph on your mobile or other platforms.

  • cmu (trained in 656x368)
  • mobilenet_thin (trained in 432x368)

CMU's model graphs are too large for git, so I uploaded them on an external cloud. You should download them if you want to use cmu's original model. Download scripts are provided in the model folder.

$ cd models/graph/cmu
$ bash download.sh

Inference Time

Dataset Model Inference Time
Macbook Pro i5 3.1G
Inference Time
Jetson TX2
Coco cmu 10.0s @ 368x368 OOM @ 368x368
5.5s @ 320x240
Coco dsconv 1.10s @ 368x368
Coco mobilenet_accurate 0.40s @ 368x368 0.18s @ 368x368
Coco mobilenet 0.24s @ 368x368 0.10s @ 368x368
Coco mobilenet_fast 0.16s @ 368x368 0.07s @ 368x368

Demo

Test Inference

You can test the inference feature with a single image.

$ python run.py --model=mobilenet_thin --resize=432x368 --image=./images/p1.jpg

The image flag MUST be relative to the src folder with no "~", i.e:

--image ../../Desktop

Then you will see the screen as below with pafmap, heatmap, result and etc.

inferent_result

Realtime Webcam

$ python run_webcam.py --model=mobilenet_thin --resize=432x368 --camera=0

Then you will see the realtime webcam screen with estimated poses as below. This Realtime Result was recored on macbook pro 13" with 3.1Ghz Dual-Core CPU.

Python Usage

This pose estimator provides simple python classes that you can use in your applications.

See run.py or run_webcam.py as references.

e = TfPoseEstimator(get_graph_path(args.model), target_size=(w, h))
humans = e.inference(image)
image = TfPoseEstimator.draw_humans(image, humans, imgcopy=False)

ROS Support

See : etcs/ros.md

Training

See : etcs/training.md

References

OpenPose

[1] https://github.com/CMU-Perceptual-Computing-Lab/openpose

[2] Training Codes : https://github.com/ZheC/Realtime_Multi-Person_Pose_Estimation

[3] Custom Caffe by Openpose : https://github.com/CMU-Perceptual-Computing-Lab/caffe_train

[4] Keras Openpose : https://github.com/michalfaber/keras_Realtime_Multi-Person_Pose_Estimation

[5] Keras Openpose2 : https://github.com/kevinlin311tw/keras-openpose-reproduce

Lifting from the deep

[1] Arxiv Paper : https://arxiv.org/abs/1701.00295

[2] https://github.com/DenisTome/Lifting-from-the-Deep-release

Mobilenet

[1] Original Paper : https://arxiv.org/abs/1704.04861

[2] Pretrained model : https://github.com/tensorflow/models/blob/master/slim/nets/mobilenet_v1.md

Libraries

[1] Tensorpack : https://github.com/ppwwyyxx/tensorpack

Tensorflow Tips

[1] Freeze graph : https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/tools/freeze_graph.py

[2] Optimize graph : https://codelabs.developers.google.com/codelabs/tensorflow-for-poets-2