Fix It Up

• 

  landing page

• 

  ideal stance

• 

  wrong posture

Inspiration

People have begun working from home during the pandemic. We've all seen our parents working on their laptops for long periods of time. It's natural that people tend to ignore their sitting posture as their workload increases and they spend more time in front of a computer. Slouching, leaning, and tilting become unconscious habits that, over time, lead to a poor physique and body condition. It can result in severe back pain, joint locking, breathing difficulties, spinal cord bone wear, and a variety of other problems. Keeping this in mind, we created the app "Fix It Up", which assists people in maintaining proper posture.

What it does

Looking at this issue that we, as working members of our society, have encountered, we decided to create an application that can alert our users whenever their posture deviates from their ideal stance. "Fix It Up" is a responsive web application that puts this concept into action using machine learning and deep AI.

How we built it

We used the TensorFlow.js PoseNet model, which tracks body parts and returns their positions in the form of (x, y) coordinates. We chose to ask the user to maintain proper posture and position their eyes and shoulders. Then, at regular intervals (1 – 2 seconds), the model checks for new positions and compares them to the first positions. It also detect arms and legs if they are visible to the camera.

Challenges we ran into

The very first challenge we ran into was how to determine the correct sitting posture. After much research we ended up with two options - Create a model and teach it to understand when a user has an incorrect posture or to ask the user to keep a correct posture, get their right position of eyes and shoulders and compare it with the initial posture. Due to time constraints we had to go for the second option. The second challenge was to recognize the various body parts. We built a model to recognize arms and legs but it ended up being highly inaccurate. So we decided to use the eyes as a reference point to determine the posture.

Accomplishments that we're proud of

Despite all the challenges we were able to figure out how to determine the posture of a person using eyes as a reference. Our model is also able to recognize parts like arms and legs with a great accuracy. Our web app is a fully functional responsive app designed using latest technologies like bootstrap, ml5.js,tensorflow.js and many more.

What we learned

We learned a lot of things while working as a team. apart from team spirit we learned how to work with multiple technologies. I gained confidence, inspiration, motivation, friends, a larger network, and mentors. I learned how to hack, how to push myself on little sleep, how to learn, how to look things up, and how to demo and sell a project.

What's next for Fix It Up

We think that Fix It Up has a good future as an extension or an independent app. It has a wide scope of applications like in proctors. If introduced in different working organizations, it can help in maintaining a healthy working environment. We could only deliver a basic working prototype of our idea due to severe time constraints. In the near future we would like to introduce substantial amount of features like introducing a sound alert for wrong posture, blurring the entire desktop instead of just the page etc.

Built With

• bootstrap
• css3
• deep-learning
• html5
• javascript
• machine-learning
• ml5.js
• p5.js
• tensorflow.js