A project to automate calculation of joint angles frop openpose JSON output.

Prerequisites:

This project requires openpose to be installed https://github.com/CMU-Perceptual-Computing-Lab/openpose
Ensure that tkinter, numpy and matplotlib are installed: pip3 install tk numpy matplotlib

Installation Instructions:

Clone the repository using git git clone https://github.com/kembr/openpose-pipeline.git or altneratively download as a zip file.

Usage:

1. After running openpose on a video file using the flag --write_json [output_folder]
   you will have a folder containing the JSON files which our project can analyze.
2. Navigate to the root directory of the project which was cloned or unzipped and run using:
   python3 src/gui.py
3. You should see the GUI pop up and be able to select the JSON folder to analyze,
   with the option to display and save output graphs.

Heplful flags for use with openpose:

--write_json <path>: specify path to save json output files.
--write_video <path>: specifiy path to save rendered video output.
--write_video_with_audio false: if input is a video, rendered video output will not be saved with audio
--disable_blending true: renders keypoint skeletons on black background instead of original video

If using hand model, change net resolution and add hand flag:
--net_resolution="-1x160" --hand

Command to run quickly on one person:
bin\OpenPoseDemo.exe --video [video_file] --write_json [output_folder] --tracking 1 --number_people_max 1

NB: the json output in each file is printed to one line, which is unwieldy. If using vim you can use :%!python -m json.tool to pretty print the json in a more human readable form, or if using VSCode or similar try using an extension like "Prettify JSON".