Maximizing workout results with real time guidance

what is fitlife.ai?

A fitness assistant that corrects your exercise posture

Fitlife uses Computer Vision and AI to provide real-time corrective feedback on the user's exercise posture during their home workouts thus, maximizing workout results. This was my capstone project during my engineering undergrad.

PROBLEM

Home workout beginners risk getting hurt from poor exercise form

Without guidance, beginners fall prey to improper posture. Only trainer or other gym-goers help in following the correct posture. Personal trainers cost too much, so most can't afford them. Plus, there's a ton of unorganized fitness content online.

solution

Provide instant audio feedback to guide their performance

We designed Fitlife.ai to enhances fitness guidance with real-time audio feedback. It analyzes exercise posture, tailors workouts to specific health conditions, and tracks progress. Here’s how it turned out!

personalization

Curated exercises playlist to match goals

Given the varied goals for exercising, we segmented users based on their current physical conditions, injuries and needs. This allows them to efficiently discover exercises that best suit their bodies, saving time and energy.

context-aware

Audio-video guide to learn while performing

Using audio-video instruction we ensured that they can focus on performing from a distance than reading about the exercise. We prioritized concise information on breathing techniques, targeted muscles, and rest intervals.

WHY DESKTOP APPROACH?

Act as a mirror to see posture from a distance

We considered a mobile approach but ultimately chose a desktop web app because its larger screen serves as a mirror for better posture viewing. And, implementing OpenCV library on mobile would be complex due iOS and Android differences.

initial protoptype

A basic demo of our body tracking model

Instead of discussing ideas, we created a prototype for knee-pushups to demonstrate our initial vision of the fitness assistant. This allowed us to get feedback on how we could improve our body tracking model.

user testing

We asked 5 people to use our prototype

reflections

Fun experiments with Protopie

The most interesting part was Protopie's 'Speak' feature that helped bring our voice feature to life. Although the internet did not solve all my questions, I was able to figure out work-arounds by putting together the basic blocks.

Navigating complexities of computer vision

Thanks to my teammates, I learned the complex technicalities of our AI model before jumping to Figma. I identified constraints clearly and centered my design around it. This helped minimize unproductive team meetings.