Posture Police

Banner
AI Generated Logo
CAD Shell Design
Blender Model
Prototype
Visual Processing Brainstorming
Activity Log
Report

Inspiration

Did you know that around 80% of all adults will experience chronic back pain? That over $50 billion is spent in the US alone to combat chronic back pain? And did you know you are more likely to suffer from depression and anxiety with poor posture? Well, no more! This project will help stop these and will allow you to achieve your full posture potential and help save the world— one back at a time!

Concept

The PosturePolice is a fun and interactive companion and reminds students to keep good posture when they're studying. The device is battery-powered and does all its processing internally, and it is connected via Wifi to the user's phone app. The PostureAI app would allow the user to enable or disable the 'posture patrol' and would provide graphics and data to the user about their posture.

Restrictions

Due to limited access to materials such as a camera and battery, we chose to build a prototype of the device, which is connected to a laptop and uses its system camera instead. This way, we were able to demonstrate the viability of this concept.

What it does

Our solution integrates advanced machine learning algorithms with real-time video analytics to accurately detect poor posture. It provides immediate corrective feedback through intuitive visual indicators and physical cues, guiding users towards healthier posture habits. By continuously monitoring and analysing user posture, our system creates personalised health insights and promotes proactive intervention.

How we built it

We employed Mediapipe's holistic machine learning model combined with OpenCV for real-time image processing, achieving robust posture detection capabilities. The backend utilises Python, integrating with SQLite databases for efficient data logging and analysis. Hardware integration with Arduino microcontrollers ensures interactive feedback through servo motors and physical alerts, enhancing user engagement and effectiveness. Throughout the process, we used AI, including Al.bot, for research and debugging.

Challenges we ran into

Implementing seamless real-time video processing while maintaining high accuracy and minimal latency presented significant technical hurdles. Additionally, integrating diverse hardware and software components required careful management of synchronisation and communication protocols. Ensuring intuitive user interaction and robust database integration added complexity, demanding iterative development and extensive debugging.

Accomplishments that we're proud of

Lots of blood, sweat, and tears went into this project, so obviously, there will be some things we are SUPER proud of:

Our response time is instant. It is a real-time posture correction algorithm that can track the poses of the human body very well.
We have made (and plan to upgrade) the bot to be as small and compact as possible, meaning that students would have an easy time using it without taking up too much space.
The cost of the hardware is also very cheap, meaning it is accessible to almost everyone out there.

What we learned

All of our team members learnt something new! From how to deal with databases or how to use computer vision. We are all excited to continue our learning and reinforce the skills we learnt in this Hackathon.

What's next for Posture Police

Our future aim for the posture police is to build the bot fully, equipping it with its own camera and possibly a speaker and an LCD display. Furthermore, we could also develop the PostureAI app, allowing users to easily check their posture stats and compare them with friends and family.