Stride

Home Page
Home Page 2
Submit Video Page
Make Coach Page
AI Coach Creation
Results page with custom coach

Inspiration

Good running form is hard to master. Finding a trained expert, nevertheless affording one is out of the question for most. Our goal was to combine computer vision, personalized AI, and gamified motivation to make elite-level coaching available to anyone, anywhere, for free.

What it does

Stride lets users record or upload a simple side-view running video, then analyzes it using MediaPipe and OpenCV to extract limb angles and detect biomechanical issues like overstriding, hip drop, or improper knee tracking. The computed data is then passed into Gemini Flash, where it’s translated into clear, actionable coaching feedback — written in natural language, tailored to the user’s specific running form, and delivered by a customizable AI coach persona.

How we built it

An idea born during downtime in cubicles, this app was built from the ground up. Utilizing Flutter's seamless cross-platform development, we used IOS emulators to create an IOS app. Flutter's modernity offers users a premium UI and a clean, responsive experience. The backend uses Python with MediaPipe + OpenCV to extract joint angles and calculate limb score. As the flagship of our application, Gemini was employed to generate personalized coach avatars as well as their summaries based on user's selections to offer a truly personalized experience.

Challenges we ran into

Integration of MediaPipe and video backend pipelines caused quite some issue for us. These challenges collectively demanded deep technical investigation, iterative debugging, and close coordination between our backend team to achieve a robust, production-ready solution. Additionally, the UCF enterprise WPA2 network was blocking our database connections due to its firewall restrictions. We resolved this connectivity issue by switching to personal mobile hotspots, which allowed our application to establish proper database connections and function as intended.

Accomplishments that we're proud of

We are proud of the end-to-end pipeline - video input, limb analysis, and custom AI coach creation all working in sync. The unrivaled UI incorporates the Gemini color palette to serve users with a soft and gentle theme on the eyes while maintaining modern design elements.

What we learned

We learned how to integrate AI and CV tools across mobile and backend environments. Additionally, UI/UX design for mobile components as well as data pipelines. We discovered the importance of network infrastructure considerations when deploying real-time applications and how to adapt quickly when technical constraints arise.

What's next for Stride

We plan to take Stride to the next level by adding voice feedback from your AI coach, implement leaderboards, and support your friends through connections.

Built With

fastapi
flutter
gemini
opencv
python
supabase

Submitted to

GemiKnights

Created by

I developed the backend and data pipeline architecture to stream video frames into mediapipe/OpenCV and finally to supabase storage.

Aiden Letourneau
I developed the main Python file that implements joint angle calculations using MediaPipe's pose estimation framework to analyze user movement and form during running videos. This component extracts key body landmarks and computes precise joint angles to provide biomechanical feedback on running form. I also helped establish the connection between the frontend and backend, implementing FastAPI endpoints to seamlessly connect our Flutter frontend with the Supabase backend.

Jack Vo
I developed the Gemini API integration creating the AI coaches image generation and the summary page summary of the data.

Brett Rosenberg
I led the frontend development by building the entire UI from scratch, designing each screen to be intuitive, responsive, and aligned with our brand vision. Throughout the hackathon, I continuously iterated and refined the interface based on user flow and feedback. In addition to UI development, I organized the overall project structure, defined key workflows for the team, and helped coordinate integration between the frontend, machine learning backend, and the Gemini AI features. This ensured we had a clear development pipeline, smooth handoffs, and a high-quality final product.

Kenneth Valladares