Subitizing with Variational Autoencoders

TL;DR

This paper demonstrates that variational autoencoders can spontaneously learn to subitize in complex natural images without supervision, encoding numerosity as a basic visual property and showing invariance to object size.

Contribution

It shows that unsupervised hierarchical neural networks, specifically variational autoencoders, can perform subitizing on natural images without supervision, aligning with biological neural properties.

Findings

VAEs can perform subitizing after unsupervised training

Networks encode numerosity as a visual property

Representations are invariant to object area

Abstract

Numerosity, the number of objects in a set, is a basic property of a given visual scene. Many animals develop the perceptual ability to subitize: the near-instantaneous identification of the numerosity in small sets of visual items. In computer vision, it has been shown that numerosity emerges as a statistical property in neural networks during unsupervised learning from simple synthetic images. In this work, we focus on more complex natural images using unsupervised hierarchical neural networks. Specifically, we show that variational autoencoders are able to spontaneously perform subitizing after training without supervision on a large amount images from the Salient Object Subitizing dataset. While our method is unable to outperform supervised convolutional networks for subitizing, we observe that the networks learn to encode numerosity as basic visual property. Moreover, we find that the learned representations are likely invariant to object area; an observation in alignment with studies on biological neural networks in cognitive neuroscience.