History: Around 15% of the individual final project's worth in the Machine Learning course i took on 2018. The approach taken could be considered almost bruteforce, since the implementation lodged to be softcoded changing only the image dimensions in the first 20 lines of code, i had yet enough experience to make a shorter version (the final implementation of the k-means algorithm took as many as 200 lines of code!), albeit it's very easy to understand and very well documented, nevertheless the results were inline with the requirements and the showcase was successful.
TASK: K-means Creating 6 clusters of random pixels as starting points , and creating 6 clusters with preset positions in for the following : (streets, trees, grass, small buildings, buildings, water bodies )
Image source : https://commons.wikimedia.org/w/index.php?curid=28250087 800px-Aerial_-east_Scarborough&_Pickering,Ontario&environs_01-white_balanced(9660041954).jpg