digitalmediaprofessionals/kerasposenet

KerasPoseNet inference

KerasPoseNet

KerasPoseNet inference example

Network used is from: https://github.com/ildoonet/tf-pose-estimation (git SHA1 ID: b119759e8a41828c633bd39b5c883bf5a56a214f, Apache 2.0 License).

To generate network inference sources and weights files use the tool:

python3 <path_to_tool>/convertor.py keras_pose.ini

To build main application on Ubuntu:

sudo apt-get install python3-opencv python3-psutil
make

To build 3rd-party helper library:

cd pafprocess
sudo apt-get install swig
swig -python -c++ pafprocess.i && python3 setup.py build_ext --inplace

To run application using Web-camera (sudo is needed to set linux console to graphics mode before drawing to framebuffer):

sudo ./run.py /dev/video0

To run application with input image:

sudo ./run.py IMAGE_PATH

To run application with multiple images (img_01.jpg, img_02.jpg, ...):

sudo ./run.py img_%02d.jpg

kerasposenet's Issues

Fix README.md: run.sh => run.py