Human-Pose-Classification

A Python and Keras implementation for classifying human arm poses

About

Trained using the InceptionV3 model on a custom image dataset of 36 different positions.

Installation

1. Requires the Openpose Python API to be properly installed
2. Place this project in the compiled OpenPose directory e.g. 'C:/openpose-master/Release/python/openpose/this_project'
3. Be sure to replace the existing openpose.py with this project
4. Download the model from ... and place it in the 'model' folder

Usage

Run 'openpose.py' (on line 416 change use_webcam to true if you want to run through a webcam or image_dir to your own directory to run through a folder of images)

Performance

InceptionV3 performance on the test data:

Statististic	Score
Accuracy	0.999

Credits

The human-pose estimation was made possible thanks to: https://github.com/CMU-Perceptual-Computing-Lab/openpose