Inspiration
LikeMike is a movie. It goes hard fr.
We've always wanted to inspire people to get out there and get active. One of the best ways that we can do that is to enable athletes to master the moves of the pros.
What it does
Can you shoot like Curry? Can you dunk like Mike? LikeMike uses advanced pose estimation and comparison techniques to allow anyone to replicate the moves of their favourite athletic icon. Simply set up your phone, start recording, and go through the motions to receive both a replication score and feedback for how to improve.
What does LikeMike do? It puts you in your bag. Check this video to see it in action:
How we built it
We used the open source MoveNet machine learning model to get pose estimation landmarks, and used cosine similarity to compare the pose with the motions of professional athletes. We also looked in to using the 3D pose estimation model MotionBert to allow people to use this app from any angle. To give feedback, we fed unstructured cosine similarity data to Respell AI to give feedback.
Challenges we ran into
We couldn't host our 3D pose estimation model because we didn't have a GPU!
Accomplishments that we're proud of
It's super satisfying and delightful to use, with actual technical MOAT rather than an API wrapper.
What we learned
Math and frontend development.
What's next for LikeMike
People will begin uploading their own bags and compete within friend groups, teammates, and even family members. Any professional or casual athlete can join in on the fun!
Log in or sign up for Devpost to join the conversation.