PostureNet

Inspiration

After persistent periods of back and neck strain during long coding sessions, we decided to investigate further on the reason behind it. As developers we all face an extremely inevitable fate: bad posture. Bad posture can lead to all sorts of chronic illnesses such as obesity, joint pain, muscle strain, and depression (just to name a few). We decided to take this matter into our own hands and decided to “have the backs” of other computer users by creating a web application that would align our posture with salvation.

What it does

PostureNet utilizes Machine Learning to improve your posture. It accomplishes this with the help of the Tensorflow and PoseNet frameworks to gather human posture keypoints, which it then runs calculations on to find specific areas of weakness in your posture.

Break Timer: If the user sits at the computer over the time they allot themselves, then we alert the user to take a break If the user is standing OR if the user is out of the picture (which assumes that the user has left the workplace), the timer will reset because the user is not currently sitting.

Statistics: Displays more detailed insights into your posture. Shoulder to hip posture - measures the relationship between shoulder and hip points generated by the PoseNet code to prevent back slouching. Head to shoulder posture - measures the relationship between Head and shoulder points generated by the PoseNet code to prevent neck strains.

Charts: Creates pie charts that displays the percentage of time spent in good posture vs bad posture and time spent standing vs sitting.

How we built it

We used the Machine Learning library PoseNet, built using Tensorflow, to estimate human pose keypoints in real time. We took these points and created algorithms to analyze what parts of your posture are good and what needs to be improved. We built the web application from scratch using html, css, and javascript. We deployed the app using Firebase.