pose_estimation_AWS

this is a pose estimation implement for AWS

This is the demo which should run on AWS to complete the image deblur task.

The code basicly from https://github.com/ildoonet/tf-pose-estimation

Install

Dependencies

You need dependencies below.

• python3
• tensorflow 1.4.1+
• opencv3, protobuf, python3-tk
• slidingwindow
  • https://github.com/adamrehn/slidingwindow
  • I copied from the above git repo to modify few things.

Install

Clone the repo and install 3rd-party libraries.

$ git clone https://www.github.com/bestat/tf-pose-estimation
$ cd tf-pose-estimation
$ 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-pose-estimation
$ cd tf-pose-estimation
$ python setup.py install  # Or, `pip install -e .`

how to use

the same as the face identification task, you can use this command,

python3 pose_estimation_on_aws.py --input_path='input.json' --output_path='output.json'

then you can get the output.json and frame.jpg.

input and output format

Input and output image format are all json file. The input json file format is

{
img:data
}

here data is the list format of one uint8 image array. For example, [500x500x3] uint8. For further details, please check generate_json_file.py as a reference. I use this file to create the input.json from p1.jpg.

The output json format is the same as the input json file format.

References

See : etcs/reference.md