Wiktoria Mróz, a Systems Engineering graduate, designed an application that analyses body posture in real time. The tool helps users maintain a correct sitting position at the computer, encourages posture correction, and promotes short stretching exercises. The project took second place in this year’s CreatiWITy competition in the engineering thesis category.

Inspired by everyday experience

The idea for the application came from long hours spent working at a computer. Wiktoria developed it as part of her engineering thesis titled “Prototype of a system for analysing and correcting sitting posture”, supervised by Dr Dariusz Gąsior.

– I wanted to create something genuinely useful. While studying and searching for a thesis topic, I spent countless hours in front of a screen. At some point, I realised the problem was right in front of me – back pain after long sessions at a laptop – explains the graduate.

She set out to design an application supporting everyday computer work – one that helps maintain proper posture and reduces strain on the musculoskeletal system.

Real-time posture analysis

The application runs in the background while working on a computer. It uses a built-in or external camera and analyses the image in real time. A key component is a pose estimation model, which detects characteristic body points such as shoulders, elbows, hands, and facial landmarks. Based on their position, the system calculates posture parameters.

The solution is based on decision rules defined by the author using literature on biomechanics and ergonomic standards.

– I established rules such as limiting head deviation from the reference position to 10 degrees, and justified this threshold in my thesis. If posture remains incorrect for over a minute, the system classifies it as improper and sends a notification – she explains. Users receive either audio or visual alerts.

Calibration module

At the beginning, the user undergoes a calibration process, allowing the system to define an individual reference posture. Future deviations are assessed against this baseline. The application also estimates the user’s distance from the camera, for example by analysing the distance between the chin and the top of the head.

– Everyone sits slightly differently and at different distances from their laptop. That’s why establishing a neutral reference position is essential – explains the author.

Motivation through movement

One of the most distinctive features of the application is its mechanism encouraging users to perform short stretching exercises. When incorrect posture is detected, the system can temporarily block the screen.

– I wanted to find a method that would genuinely motivate users to stretch tense muscles. That’s why I introduced a screen-lock mechanism. Background processes continue to run – it’s essentially an overlay that temporarily blurs and blocks the screen – says Wiktoria Mróz.

To unlock the screen more quickly, users can perform three short exercises without leaving their workstation:

• stretching the neck muscles to the left,
• stretching the neck muscles to the right,
• stretching the shoulder girdle by raising the arms overhead

The system uses a machine learning classification model to verify whether exercises are performed correctly.

– The model checks if the movement is actually being performed. It’s not easy to cheat the system – if the movement is interrupted, the timer resets – she explains.

To ensure reliable detection, the model was trained on four data classes: three exercises and an additional “undefined posture” category.

– I recorded each exercise in thousands of frames, under different conditions – closer and further from the camera, with varied backgrounds. This allows the model to work effectively across different users – she explains.

A system designed to support, not disrupt

The goal of the project was to create a tool that supports users without interfering with their workflow.

– I didn’t want the application to be intrusive. Users can choose between sound alerts, visual notifications, or screen lock with exercises. If they prefer not to exercise, they simply wait longer for the screen to unlock – emphasises the author.

The application has been designed as an open-source solution, with plans for cross-platform support and broader accessibility in the future. Its aim is not only to improve posture but also to build healthy habits during computer work.

– It’s not just about correcting posture, but about motivating users to maintain it over time and remember to take short exercise breaks – concludes the graduate.