Pose Correctinator

A real-time posture-correction example for the Luxonis OAK-4. Runs the neural-net on the camera and alerts your laptop when posture is incorrect.

Prerequisites

• An OAK-4 camera with network access (PoE or Ethernet)
• Python 3.7+ on both camera and laptop
• On the camera, install:

  pip install depthai depthai-nodes numpy

• On the laptop, install:

  pip install simpleaudio

• oakctl CLI installed and configured

Setup & Usage

1. Clone this repository

   git clone https://github.com/TamseSaso/pose-correctinator.git
   cd pose-correctinator

2. Configure your laptop IP
   Open main.py and update the LAPTOP_IP constant to your laptop’s local network address:

   LAPTOP_IP = "192.168.1.100"  # replace with your laptop’s IP

3. Adjust posture detection thresholds
   In main.py (around line 152), locate the evaluate_posture() function’s angle_threshold and dy_threshold parameters. Tweak these values to calibrate the sensitivity and match your ideal posture.

4. Start the listener on your laptop
   In one terminal, run:

   python posture_alert_listener.py

5. Deploy & run on the OAK-4
   In another terminal (still in the repo root), run:

   oakctl app run .

Now, whenever the OAK-4 detects incorrect posture, it will send a UDP ping to your laptop and you’ll hear the alert sound.