This Github repository contains code to run the k-means algorithm for image compression and create corresponding visualizations.
All of the required libraries can be found in the requirements.txt file.
The basic_viz.py file contains code that performs k-means in 2D space. It may be worthwhile to reference this code first if you are new to k-means. The iterations_4.gif provides a visual explanation of how k-means works on a simple 2D example.
The main file to run k-means for image compression is k_means.py. Simply modify the filepath in parser.add_argument('--large_path', default='cuttlefish.jpg', help='Path to large image') to match your jpg file. The code in that file allows for users to create not only the compressed output image but also visualizations of the entire process, including the progression of the centroids.
Below are some examples of the potential use cases:
