SlouchLess

SlouchLess is a lightweight desktop app that watches your posture through your camera and nudges you when you start slouching.

How it works

• Uses RTMPose pose detection to track your neck, shoulders, and head position in real time.
• Lets you pick a pre-trained posture-detection model (neural net, SVM, or random forest) at startup, or fall back to a short guided calibration (sit upright, then slouch on cue) that measures your own upright vs. slouched posture to set personal alert thresholds. You can switch between them mid-session from the "Switch Model" button on the video window.
• Adapts its detection method to where your webcam actually sits - above your monitor or below it - since that changes which signals are reliable.
• Pops up a (mildly judgmental) reminder once it's confident you've been slouching for some time.

Just want to run it?

Download the prebuilt binary for your OS from the Releases page.

Running from source

Requirements:

• Python 3.11+ (developed/tested on 3.12).
• A webcam or built-in camera
• Linux only: a few system packages not covered by pip:

  sudo apt install python3-tk libgl1 libglib2.0-0 libsm6 libxext6 libxrender1 pulseaudio-utils

Setup:

python -m venv venv

# Windows
venv\Scripts\activate

# Linux
source venv/bin/activate

pip install -r requirements.txt
python main.py

Building the exe/binary yourself

After following the "Running from source" setup above, just add PyInstaller and build:

pip install pyinstaller
pyinstaller --noconfirm SlouchLess.spec

The build output lands in dist/

Notes:

• On Linux, PyInstaller also needs the binutils package (objdump) installed system-wide: sudo apt install binutils.
• Supports two camera positions, above or below the monitor. Results in other setups aren't verified.

Acknowledgments

Directing architecture and posture-detection logic, debugging cross-platform issues, and tuning calibration by hand. Code built iteratively with Claude Code