Gaussian Mixture Models Tutorial

Introduction

This repository is the accompanying software for my mathematics and programming tutorial paper for Gaussian Mixture Models. See https://alexhagiopol.github.io/content/gmm_tutorial.pdf.

Installation

Below commands tested on Ubuntu 22.04:

git clone https://github.com/alexhagiopol/gmm
cd gmm
sudo apt install git
sudo apt install pip3
sudo apt install python3-tk
sudo apt install python3-pil python3-pil.imagetk
pip install -r requirements.txt

Usage

Command line parameter definitions:

-h, --help      Show help message.
--first-image   Path to image file. Must be specified.
--second-image  Path to image file. May or may not be specified.
--components    Number of components in the mixture of Gaussians. Must be specified.
--iterations    Number of Expectation Maximization iterations. Must be specified.

Example Commands

Segment a single image:

python3 gmm_segmentation.py --first-image=example_data/beyonce.jpg --components=3 --iterations=8

Segment the difference between a pair of images (reproduce Figure 7 in paper):

python3 gmm_segmentation.py --first-image=example_data/image_pairs/2_background.png --second-image=example_data/image_pairs/2_foreground.png --components=2 --iterations=6 --subtraction-threshold=5.0

Run the code with no visualization for profiling purposes:

time python3 gmm_segmentation.py --first-image=example_data/beyonce.jpg --components=3 --iterations=8 --visualization=0

Using Precompiled C++ Functions

This project has custom-implemented C++ functions to increase performance which is especially noticeable on larger resolution datasets like the example image church.jpg. These functions require (a) submodules to be cloned, (b) existing installation of a C++ compiler invokable by make and CMake invokable by cmake, (c) running the precompilation script:

git clone --recursive [email protected]:alexhagiopol/pybind_examples.git
cd gmm
python3 build_precompiled_functions.py

Once build_precompiled_functions.py finishes successfully, the GMM segmentation can be re-run by using the --precompiled-num-threads parameter to specify a max number of threads to use for the precompiled functions.

Run the code with no visualization for profiling purposes:

time python3 gmm_segmentation.py --first-image=example_data/church.jpg --components=3 --iterations=8 --visualization=0 --precompiled-num-threads=12

Example results:

Single image segmentation into 3 components (approximately "white", "black", and "grey") over 8 iterations: