DeepSetNet: Predicting Sets with Deep Neural Networks

TL;DR

DeepSetNet introduces a neural network approach for predicting sets of entities, effectively handling variable sizes and order invariance, with applications in image classification, counting, and detection.

Contribution

It defines a set likelihood and a cardinality loss within deep learning, enabling set prediction tasks to be learned directly from data.

Findings

Outperforms existing methods on image classification, counting, and detection datasets.

Effectively models set cardinality and permutation invariance.

Applicable to multiple computer vision tasks with improved accuracy.

Abstract

This paper addresses the task of set prediction using deep learning. This is important because the output of many computer vision tasks, including image tagging and object detection, are naturally expressed as sets of entities rather than vectors. As opposed to a vector, the size of a set is not fixed in advance, and it is invariant to the ordering of entities within it. We define a likelihood for a set distribution and learn its parameters using a deep neural network. We also derive a loss for predicting a discrete distribution corresponding to set cardinality. Set prediction is demonstrated on the problem of multi-class image classification. Moreover, we show that the proposed cardinality loss can also trivially be applied to the tasks of object counting and pedestrian detection. Our approach outperforms existing methods in all three cases on standard datasets.