CMU uses the OpenPose model, while OpenCV integrates the pre-trained OpenPose model in its new deep neural network (DNN) framework. The entire code block can be downloaded from the following GitHub page. This model uses the TensorFlow example instead of the Caffe model that was originally used by the OpenPose authors, can be found at https://github.com/quanhua92/human-pose-estimation-opencv.
OpenCV's OpenPose code can be executed in a Terminal using the following command:
python openpose.py --input image.jpg
To get started using your PC's web camera, just type the following in your Terminal:
The following image shows the implementation of OpenPose for an image of a soccer player:
The algorithm ...