We were inspired by some gym correction videos before.

What it does

It use webcam to capture the poses of the user, check if they are "standard" enough, and give voice reply to the users.

How we built it

We used OpenPose to extract pose information. Then we used machine learning to build a model that represents the "standard" pose. The whole system is deployed on cloud server.

Challenges we ran into

We met difficulties in putting all code onto the cloud. Also we spent much time labeling for machine learning data. Putting multiple parts of the codes (machine learning, web app, etc.) together to work as a system was also a challenge we faced.

Accomplishments that we're proud of

Our codes managed to transmit all the video data and analyze them. It even connect to soundhound's voice-to-text API.

Built With

Share this project: