Visualização de obras pictóricas depois da análise de clusterização

Abstract

There is a growing need to develop more intuitive visualization tools to comprehend the results of machine learning algorithms. This study presents the implementation of image clustering visualization, using results obtained from a previous work that examined the dimensionality reduction of 304 Impressionist artworks with PCA and t-SNE, and clustering methods such as k-means and hierarchical clustering. The application features the plotting of images on the graph instead of abstract points, as well as the application of zoom when interacting with the data. The objective of the application is to propose a more evident visualization for image clustering. The visualization was developed using the Python programming language and its libraries, including matplotlib, annotationbBox, and Ipywidgets. Finally, the results of the visualizations are demonstrated along with interaction with the images.