This repository contains a computer vision Python project developed by #team1. The project idea was brought forward by Amal Tili from Pyladies Tunis. The presentation of the project can be found in the YouTube Video and the Powerpoint presentation
During the COVID19 pandemic, many people were unable to attend recreational centres and fitness centers, and needed to rely on home tools. Furthermore, people experienced an increased sense of isolation and loss of motivation while working out at home. Some people might turn to a personal trainer who encourages them to continue physical training . In order to avoid further contact with strangers, and to ease social distancing measures, we are adapting a Python program that detects body pose with the purpose of helping people to count the repeated body exercises at home.
Our adaptation would help a person who is exercising count the number of repetitions of a given exercise they are doing, and also track their body posture as they exercise. Tracking body posture data can then be used to examine the movements that might cause injury.
This repository contains the jupyter notebooks and scripts for physical exercise classication and counting
From all team members, videos of squat and bicep curls were collected and stored in this folder.
From each video, 50 frames containing pose joints were retrieved for each observation. The pose joint estimation was implmented with Mediapipe Llbrary.
All the observations were stored in .csv format. The .csv file is then used for training
- Normalize joint coordinates relative to center point of the body.
- Structure input array to represent video frames per exercise repetiion
- Build LSTM deep learning model, train and test.
- Package model and create modular preprocessing package.
- Create main script to ingest new video and predict with saved model.
A simple webpage was implemented to demo the counting of the exercises video from Webcam
A new exercise video is processed and the exercise type is predicted by the developed LSTM model. A counting.py code is to count the number of moves for each type of exercise: squats and curls
| Name | Role/Tasks | Github | ||
|---|---|---|---|---|
| Nasreen Mohsin | Team Lead / Documentation, Code sanity check, Algorithm development, Project management | nasreenpmohsin | [email protected] | https://www.linkedin.com/in/nasreen-mohsin-08210419/ |
| Srishti Yadav | Team Co-Lead, Verify code reproduction, Project management, Web development | copperwiring | [email protected] | https://www.linkedin.com/in/srishti-yadav/ |
| Sami Ma | Mentor | KamiCreed | [email protected] | https://www.linkedin.com/in/sami-ma-6b616d69/ |
| Ketian Bai | Literature Review, Counting code, Presentation | ketianBai | [email protected] | https://www.linkedin.com/in/%E5%8F%AF%E7%94%9C-%E7%99%BD-b43a88200/ |
| Momo | Data Feature Extraction | momoueda | momoueda [email protected] | https://www.linkedin.com/in/mueda |
| jason | Data Feature Extraction | yjc2 | ||
| Anna Jose | Data Feature Extraction, Presentation | annacjose | [email protected] | |
| Chloe Zhou | Building/training Model | Chloe-Zhouu | [email protected] | www.linkedin.com/in/ming-chloe-zhou |
| Austin Ngo | Building/training Model | austyngo | [email protected] | www.linkedin.com/in/austinngo/ |
Main page: https://vancouverdatajam.ca/
| Date | Action item |
|---|---|
| Sep 13 - 17 | Participants are let in Discord, teams are formed |
| Sep 18 | Workshop day! Keynote: Making AI responsible with May Masoud |
| Sep 19 | Project statements are released |
| Sep 19-24 | Teams may work asynchronously (limited help desk support) |
| Sep 25 | Keynote talks, help desk support provided during the day, project submission deadline, career panel. See speakers |
| Time | Action item |
|---|---|
| 8:00 - 8:10 | Land acknowledgement, opening remarks |
| 8:10 - 8:40 | Keynote: Role of Statistics in Data Science: Applications in Biomedical Sciences with Prof. Jemila Hamid |
| 8:40 - 9:10 | Keynote: How to use the tools of data science to benefit Indigenous peoples and organizations with Hannes Edinger |
| 9:10 - 9:30 | Keynote Q&A |
| 9:30 | Help desk opens up, teams work on their project |
| 9:30 - 10:00 | Teams brainstorm tasks for the day |
| 12:30 - 13:00 | Team check in: share exploratory analysis results |
| 15:30 - 16:00 | Team check in: teams discuss presentation format and preliminary results |
| 16:00 - 16:45 | Teams prepare their 5-10 minute presentation, teams ensure all code is documented and stored in GitHub |
| 17:00 | Project video submission deadline |
| 17:30 - 18:30 | Project videos released on YouTube. Vote for your favourite team demo! |
| 18:30 - 20:00 | Career panel |
| 20:00 - 20:30 | People's Choice Award presented. Closing remarks |