PosturePal

PosturePal Logo
PosturePal App Home

PosturePal

Inspiration for our project: Good posture is essential for long-term health, yet many people struggle to maintain it, especially while working at a desk for extended periods. We wanted to build a solution that provides real-time feedback and actionable insights to help users develop better posture habits.

🎯 What It Does

PosturePal analyzes a user's posture using their webcam and provides instant feedback. It detects key body landmarks, calculates posture metrics, and assigns a score based on alignment. If poor posture is detected, the system generates a summary and personalized improvement strategies using AWS Bedrock and Mistral AI (Mistral Large 24.02). Users can also track their posture history and trends over time with visualizations.

🛠️ How We Built It

We developed PosturePal using a full-stack approach:

Frontend: Built with React, TypeScript, and Tailwind CSS for a responsive and intuitive user experience.
Backend: Flask API handles authentication, posture data processing, and integration with AWS services.
Machine Learning: TensorFlow.js and MediaPipe power real-time posture analysis directly in the browser. Also trained our own Neural Network on dataset.
AWS Bedrock: Runs Mistral Large to summarize posture data and generate personalized improvement strategies.
Database: Supabase (PostgreSQL) stores user data, posture scores, and session history.

🚧 Challenges We Ran Into

Fine-tuning the posture detection model to minimize inaccuracies was a key challenge. Ensuring real-time performance while integrating TensorFlow.js, MediaPipe, and AWS Bedrock required significant optimization. Additionally, generating meaningful and concise recommendations using Mistral involved testing different prompt structures.

🏆 Accomplishments That We're Proud Of

Successfully integrating real-time posture tracking with machine learning.
Implementing AWS Bedrock with Mistral to provide personalized posture improvement strategies.
Building a seamless user experience with authentication, tracking, and analytics.
Implementing Google OAuth for user sessions and secure login experience.

📚 What We Learned

This project deepened our knowledge of real-time computer vision, AI-powered summarization, and optimizing machine learning models for web applications. We also gained hands-on experience with AWS Bedrock and deploying scalable AI-driven solutions.

🚀 What's Next for PosturePal

We're looking to refine our posture scoring system with more robust data and feedback. Additionally, we plan to improve Mistral’s recommendations, expand device compatibility, and explore mobile support for a more accessible experience. PosturePal is more than just an app—it’s a lifestyle change powered by AI. 🚀