Yoga Pose Classification App

COSC 490 Final Project due 11 April 2022

Authors: Edouard Eltherington, Veronica Jack, Logan Parker, and Khai Hung Luong

Table of Contents

• General Information
• Language and Modules
• Setup
• Walkthrough
• Citations

General Information

Please view our project proposal and final report to learn more about our application's development.

Language and Modules

This application was created with the following language and module versions:

• Python 3.9.7
• streamlit 1.8.1
• tensorflow 2.7.0
• keras 2.7.0
• numpy 1.21.3
• opencv-python 4.5.5.64
• scikit-image 0.19.2
• urllib3 1.26.7

Setup

Important: This version is compatible both on Windows and MacOS systems. Please note that this version has reduced performance due to Openpose integration on MacOS computers. If you are using a Windows computer is it highly recommended that you download this repo from the ‘window-only’ branch.

1. Before you get started, please ensure you have Python 3.7, PIP, and the modules listed above installed.

2. Download and Prepare OpenPose

   1. Download openpose-1.7.0 (portable version of OpenPose)
   2. Extract the folder openpose and move it into the Yoga-Pose-Classification app (note: put inside not on the same level).
   3. Download the caffemodel and place the file in "openpose/models/pose/body_25/"

3. Run the application, which will open the app your default web browser.

   cd src
   streamlit run main.py
   

Walkthrough

1. Home Screen

2. Select a pose

3. Attempt yoga pose, take a photo, and process image

Citations

OpenPose models are taken from CMU-Perceptual-Computing-Lab/OpenPose for study purpose, link to the source repository: https://github.com/CMU-Perceptual-Computing-Lab/openpose