Featuring the topology with the unsupervised machine learning

TL;DR

This paper demonstrates that an unsupervised autoencoder can effectively capture and preserve the topological features of line drawing images, such as winding numbers, enabling topology-aware image clustering.

Contribution

The study introduces an autoencoder model that preserves topological information in images, advancing unsupervised learning for topology recognition.

Findings

Over 90% accuracy in retaining topological information

Unsupervised features enable topology-aware clustering

Model distinguishes primitive elements in line drawings

Abstract

Images of line drawings are generally composed of primitive elements. One of the most fundamental elements to characterize images is the topology; line segments belong to a category different from closed circles, and closed circles with different winding degrees are nonequivalent. We investigate images with nontrivial winding using the unsupervised machine learning. We build an autoencoder model with a combination of convolutional and fully connected neural networks. We confirm that compressed data filtered from the trained model retain more than 90% of correct information on the topology, evidencing that image clustering from the unsupervised learning features the topology.