This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
I built SportFit, a gamified, AI-powered web platform designed to motivate people to step away from their screens, get active, and explore their surroundings.
SportFit gets people into the real world through features like AR Campus Quests, which gamify real-world walking by assigning location-based missions, and FitCoin Rewards, which incentivizes users to work out in groups using GPS. It also includes an AI Form Guard to ensure users are exercising safely and effectively wherever they are.
This platform is for anyone looking to build healthier habits, from students exploring their campus to fitness enthusiasts wanting real-time form correction. Crucially, it includes a dedicated PW Care and Accessibility section with an AI Voice Coach, ensuring that fitness and movement are accessible to physically challenged individuals and those relying on screen readers or voice navigation.
Demo
Code
https://github.com/Shivam-Sharma2009/sportfit-innovexa-the-techlets
How I Built It
SportFit is built using vanilla HTML, CSS, and JavaScript, prioritizing a lightweight, accessible, and responsive user experience.
To bring the project to life, I integrated several open-source technologies:
- AI Form Guard (Open-Source Local Inference): I utilized the open-source Human.js library, which runs TensorFlow.js pose-detection models entirely locally in the browser. This powers a real-time exercise counter and form corrector (for push-ups, squats, lunges, and curls) using local webcam inference without any server latency.
- AI Voice Coach: I leveraged the browser's native Web Speech API and SpeechSynthesis to create a hands-free, interactive voice coach that guides users through workouts and allows them to navigate the platform using voice commands.
- Gamification & Logic: Custom JavaScript handles the logic for the AR Campus Quests and Fatigue Detection modules.
Why Does Open Innovation Matter?
Open innovation is the backbone of SportFit. By using open-source libraries like Human.js that execute models locally, I was able to build a privacy-first application. Users' webcam feeds never leave their devices, ensuring complete privacy during their workouts.
This local inference approach also means zero network latency, which is critical for real-time form correction. A closed API would have introduced lag, required ongoing subscription costs, and raised massive privacy concerns regarding video data. Open-source AI made it possible to deliver a premium, responsive, and secure fitness coaching experience directly in the browser.
My Agent Session
Prize Categories
- Week 1: Touch Grass
- Maintainer/Contributor (if applicable)
Top comments (0)