Inspiration

One of the leading causes of worker injuries is ergonomics, where factory workers, construction workers, and even regular staff put themselves at risk when they move heavy equipment or materials. Since our team has experience in machine learning, we wanted to see if there was a way we could harness the power of computer vision and machine learning to detect these high risky poses and alert workers when they perform them so they can continue to live happy healthy lives.

What it does

Given permission to your computer's camera, the web app takes the live feed and identifies when you are in a potentially dangerous pose which can injure you. When it detects this, the bar on the page turns red.

How we built it

We used PoseNet from ml5.js for pose detection and we used p5.js for marking the specific keypoints and drawing the skeleton. We used machine learning for detecting potentially dangerous poses and notifying the person.

Challenges we ran into

One of the biggest challenges we faced was communication across time zones. Each of our team's members was located in a different time zone, which made communication difficult because at any given moment, half the team was sleeping. We were able to work around this by posting updates before we logged off so that everyone would know where we were at. Another issue we faced was simply starting. When looking at the challenge, we were given just a set of data and told to go forth and conquer. As a result we had to spend a lot of time deciding on what exactly we wanted to do and how when building our model. One of the first decisions we made was Python verses JavaScript. We went with JavaScript because we decided our ultimate goal was to have some way of logging into a website/server so you could remotely monitor a livestream from an outside location.

Accomplishments that we're proud of

It works! Even after facing a lot of challenges, and starting pretty late, we were able to make this project.

What we learned

We learnt that we should have a proper plan of action and designs when starting out. We ended up very disorganized at times over communication, but we learned to manage our time. We also learned how to do machine learning in JavaScript since we wanted to deploy this on a website.

What's next for SafePose

  • Adding hardware! We would like to deploy our model on a remote camera which can then beep or light up to alert people when they are in an unergonomic pose.
  • Authentication and login! We want our users to only have access to their cameras.

Built With

Share this project:

Updates