NOTE: Our project video talks about our goal functionality for DancerFlow as opposed to what we have already implemented.

Inspiration

Since the COVID-19 pandemic began, many forms of entertainment, including dance, are harder to enjoy in person. At the same time as many dancers are unable to practice at their studios, there has been a rise in viral dances online, particularly on TikTok. We sought to address how we might deliver a personalized dance learning experience to dancers at various levels staying at home.

What it does

Our mission is to empower dancers at any level by giving them immediate feedback so that they have the tools to see progress and have fun. With this in mind, we created DancerFlow. DancerFlow allows users to pick from our selection of dance moves. From there, the user will be prompted to perform the move and will be given a visual representation of how accurate their performance of the move is.

How we built it

In order to give users feedback, we used PoseNet, which is an open-source vision model used for estimating poses. It works by taking in either a video or photo and estimating where key body parts are. These can be body parts like the head, shoulders, elbows, and hands. Then we compare a video to your pose on a camera using a PoseNet overlay, which helps the user determine how accurate their movement or pose was.

Challenges we ran into

A main challenge we ran into was using PoseNet to get user-inputted picture or video submissions and being able to use those to compare poses. We also couldn't get webcam running when we hosted our project on repl.it even though it was working on our local machine.

Accomplishments that we are proud of

Our initial thought of how to compare two poses did not work out. So, we are very proud of figuring out a way to work around that and successfully compare two images or videos to each other at the same time.

What we learned

Along the way, we learned many things. For one, we learned how to add a carousel feature to our homepage. As previously mentioned, we also learned how to compare two pictures or videos to each other.

What's next for DancerFlow

First and foremost, our next step with DancerFlow is to quantify accuracy results and to add them directly to the result page. Additionally, we hope to be able to expand DancerFlow to include user input for dance moves and submissions. We also hope to expand our selection of moves to choose from.

Built With

Share this project:

Updates