qts8n/human-seg-mediapipe

Human segmentation on video/stream using Mediapipe

human-seg-mediapipe

Human segmentation on video/stream using Mediapipe

Requirements

• Python v3.9 or newer (works on v3.12 as well)
• vc_redist latest (windows only)
• CUDA v10.2 or newer (optional)

The script uses background and foreground animations stored in their dedicated directories in the assets directory.

Every directory should contain animation frames in .png format.

The script should automatically determine your default camera and connect to it.

The script will automaticall use your Nvidia GPU if you have one and all the necessery drivers are installed.

Setup (Linux / MacOSX)

Setup virtual environment and install dependencies:

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Run (Linux / MacOSX)

python main.py

Setup and Run (Windows)

init-camera-1.bat

Keyboard Controls

• Q - quit the application.
• F - enter fullscreen mode.
• N - enter normal mode.