Featuring the topology with the unsupervised machine learning
This work addresses the challenge of automatically extracting topological properties from images, which is important for computer vision and pattern recognition, but it appears incremental as it applies existing autoencoder methods to this specific domain.
The paper tackled the problem of characterizing images by their topological features, such as winding degrees, using unsupervised machine learning, and achieved over 90% accuracy in retaining topological information through an autoencoder model.
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.