PosePal: A Virtual Exercise Assistant

Inspiration

Many people struggle to exercise regularly due to limited access to gyms, high costs of personal trainers, fear of exercising in public, or lack of guidance on how to start.
We wanted to create a solution that makes fitness accessible, personalized, and effective for everyone, regardless of experience, location, or physical ability.

What it does

PosePal is an interactive, customizable virtual trainer that:

Creates workout routines tailored to your needs and goals
Provides feedback on performance and safety
Adapts exercises for muscle toning, fat burn, joint health, or accessibility needs

It has two main components:

Webcam – Tracks movement and analyzes poses for proper form
Chatbot – Responds to user prompts about exercise habits and goals with personalized guidance

How it Works

User Prompt – The user types in a prompt detailing their current workout habits, needs, and constraints
Personalized Workout – The chatbot generates a customized workout regimen
Video Capture – The user performs exercises in front of their webcam
Feedback from AI – The AI compares movements to its data and provides improvement and safety feedback

How we built it

Frontend: HTML/CSS/JavaScript for user interface and webcam integration
Backend: FastAPI + Python to handle chatbot logic, pose analysis, and workout generation
AI Components: Pose detection models for form analysis and a chatbot for personalized guidance
Workflow Integration: Combined webcam tracking and AI analysis with the chatbot to create a smooth, real-time user experience

Challenges we ran into

Ensuring accurate pose detection across different body types and exercise forms
Providing real-time feedback quickly enough to be useful
Balancing general exercise knowledge with highly personalized routines
Designing an interface intuitive for users of all ages and technical comfort levels

Accomplishments that we're proud of

Delivering the benefits of a personal trainer to users who may not afford one
Making exercise accessible to people with disabilities or specific physical needs
Providing data-driven education to improve users’ fitness knowledge
Successfully combining webcam tracking, AI feedback, and chatbot interaction

What we learned

Real-time pose tracking is challenging and requires handling edge cases carefully
Personalization significantly improves engagement and exercise effectiveness
Integrating multiple AI components requires thoughtful design for smooth UX
Accessibility and clear feedback are essential for encouraging consistent exercise habits