There are many different types of dances in the world, and through the use of technology, people are able to discover the many different types. If they wanted, instead of learning of them, they could also learn them by following dance tutorials. However, many people stop or give up at the phase between discovering a cool choreography and learning it correctly. Not everybody is able to analyze their own moves and compare them with the choreography, so that is the problem that we wanted to solve. We wanted to create a place and environment in which having a place to learn how to dance or follow a choreography the RIGHT way is possible.
What it does
By inputting a recording of your dance and another video of the official choreography of the dance, M(L)ove compares both of the videos to show you the similarity percentage for the dances so you could improve as the dancer.
How we built it
- Our machine learning model was posenet which analyzed the movement of the dancers
- aws to store videos in which the machine learning model would do it's magic
- bootstrap for website
Challenges we ran into
Some challenges that we ran into throughout the hackathon were getting a working pose estimation model that would work without a GPU. Also being able to connect the backend to the frontend and making the whole video comparison machine to recognize the video. As well as storing the uploaded video files.
Accomplishments that we are proud of
When we fixed the problem of the machine in the backend finally being able to recognize the videos that were uploaded was a great moment.
What we learned
For the backend, we had to learn Tensorflow.js and scope the structures of Flask from Python.
What's next for M(L)ove
We could create a leadership board that would show the highest scores that users have gotten which would create a fun competition for the users and increase motivation to do better.