RepLex | Devpost

Home Screen
Pose detector
Creating workouts

Inspiration

Many students in physical education classes struggle with accurately tracking their exercise reps, whether due to cheating, guessing, or simply losing count during workouts. It's estimated that a significant portion of students either overestimate their performance or fail to meet their exercise goals due to these inaccuracies. This common issue undermines the effectiveness of physical education programs, impacting students' health and fitness progress. Inspired by this challenge, our project aims to develop a tool that ensures accurate tracking and reporting of exercise activities. RepLex is designed to help students focus on real performance improvement by providing clear, objective measures of their workout achievements.

What it does

The user can first create their own personal workout and customize it to their own liking. Then, they can choose any workout they created in order to do the selected exercises. With the user filming themself while performing the exercises, RepLex counts the user's remaining reps and sets. As such, the user is free of the hassle of counting. Moreover, gym teachers don't have to be suspicious of students cheating because the model never lies.

How we built it

We built it using React Native, a cross platform framework that is written in React like code in javascript but that compiles to native iOS and Android code. Movenet, a pre-built tensorflow pose detection model from Google is also used in order to detect the user's poses.

Challenges we ran into

React native and mobile development does not have the general community support that Web has. As such, finding solutions on online forums were much harder than if we were making a web-app. Moreover, the tensorflow-react-native library that we were using was outdated. As such, many conflicts arose between node packages of the app.

Accomplishments that we're proud of

Although our team was already familiar with React, it was still very hard to manage state variables in React Native. As a team, we are proud of learning React Native and Mobile Development in such a short amount of time in order to deliver a tangible product useful for all Gym faculties.

What we learned

We learned React, React Native, Expo projects, tensorflow-react-native and Computer vision.

What's next for RepLex

Currently, there are not many featured exercises that can be correctly counted. The next step would be to add the related conditions to new exercises so that the model recognizes them as well.