Evolving Images Using Transparent Overlapping Polygons

Our project attempted to use machine learning to recreate or redefine features of an image using an arrangement of transparent overlapping polygons. From genetic algorithm[1] and hill climbing[2], we came up an implementation that starts by generating a random sequence of polygons then iteratively mutating the sequence (slightly modifying a random attribute of a random polygon), incrementally building on mutations that yield results that are closer to a target image[3]. In the context of evolving images using polygons, we learned and explored the balance between visual appeal of the generated images and the efficiency of the implementation.