Atenta: Fixing Posture and Improving Health with Laptop Webcams

Ethan Weber
3 min readJan 19, 2020

--

Detecting Posture with Atenta: Green is good posture, and red is bad posture. The program runs entirely in a browser tab and in the background. https://atenta.io/

When my grandfather passed away due to cancer, I became obsessed with keeping those around me healthy. This was in fourth grade, and since then, I’ve held my family members accountable to eating healthy and working out. I’ve annoyed my mom all too often trying to suppress her sweet tooth, but she thanks me occasionally. :)

Although always trying to stay healthy, it wasn’t until sophomore year at MIT that I realized I was in a unique place to spread this passion to others. This is when I started prototyping Atenta: a webcam coach for a healthier lifestyle.

It started off called Countenance, named after my grandfather and his support for me. Every time I’d visit, he’d give me a blessing, saying “…may The Lord lift up His countenance upon you, And give you peace.”

Ever since middle school, I’ve been a builder. Starting with Arduino and Raspberry Pi, I built projects that include a hovercraft, quadcopter, surveillance system, many robots, augmented reality apps, and more (https://ethanweber.me/). Looking back, I realize the common theme for all my projects is tangible and/or visual interaction. Most recently, I’ve taken a huge interest in computer vision — getting computers to understand visual aspects of the world.

Sophomore year, I finally connected the dots between my passion for health and my passion for computer vision. As a computer science student, I spent almost all my waking hours in front of my laptop — hunched over to read and code. If only I had a personal coach to tell me when I started to slouch.

My bad posture was visually noticeable, but it was also decreasing my productivity and causing severe hand pain. During my coding projects, I’d have to take breaks away from the keyboard to rest my hands. I bought hand strengthening equipment, watched YouTube stretching videos to prevent Carpal Tunnel Syndrome, and read ergonomics literature. However, I couldn’t find a clean way to solve my problem without wearing a physical device. With around 100 million desk workers in the US and $350 spent annually per employee due to lower back pain, I knew I wasn’t alone.

Then it hit me: my laptop webcam is conveniently placed to see and track my posture with computer vision algorithms.

After realizing the potential of the webcam, I became excited to explore ways to fix habits besides posture: touching face, biting nails, focus tracking, fatigue monitoring, water drinking habits, and more. However, I started with posture and built out an initial prototype (https://ethanweber.me/documents/posturepal.pdf) with my good friend Moin Nadeem.

Initial feedback for Atenta was promising, although privacy was often brought up as an issue. To mitigate privacy invasion, I landed on an elegant solution that keeps all computer vision code in the browser thanks to TensorFlow.js. Furthermore, I built a visual product that shows you what exactly the computer is “seeing” with skeletal overlays and colors.

Along the journey, I’ve gone through many product iterations, added and removed features, received funding and mentorship from MIT Sandbox, and interviewed with Y Combinator.

The current site is at https://atenta.io/, and a demo video is shown below. Simply log in, “Calibrate” for a desired skeleton position, and get notified when you deviate from your target. You can minimize the tab and go about your work, recalibrating as needed. Make sure to have notifications turned on.

You can either log in with a Google Account or Email/Password. Feel free to use the following test account too. Email: test@atenta.io, Password: password.

Atenta Product Demo: A video describing some of the features of Atenta and how to setup posture feedback.

I’m confident a webcam health coach will exist for those who want it, and I hope to help build this future. Additional features I plan to add include history tracking, dimming the screen as an alternative to pop-up notifications, and tracking other health habits. In fact, I use a face touching feature locally to prevent spreading germs and inflaming acne.

If you’d like to collaborate, please reach out to me at ejweber@mit.edu. If you or a friend could benefit from Atenta, please encourage me by sending an email my way. :) I’m happy to build out custom features for those who would really benefit.

Big shoutout to Moin Nadeem for helping with early prototypes, and to Jordan Wick for recommending to start with a posture product.

--

--

Ethan Weber

I’m passionate about helping others with technology projects. Particularly excited by robotics, computer vision, and augmented reality. https://ethanweber.me/