Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects

TL;DR

This paper introduces a unified deep learning framework for simultaneous salient object detection, ranking, and subitizing, addressing the ambiguity in defining saliency among multiple observers and improving performance across various metrics.

Contribution

It proposes a novel hierarchical deep learning approach that models relative saliency and subitizing within a unified network, advancing beyond prior isolated methods.

Findings

Outperforms previous methods on all evaluated metrics

Effectively models relative saliency and object ranking

Addresses the subitizing of multiple salient objects

Abstract

Salient object detection is a problem that has been considered in detail and many solutions proposed. In this paper, we argue that work to date has addressed a problem that is relatively ill-posed. Specifically, there is not universal agreement about what constitutes a salient object when multiple observers are queried. This implies that some objects are more likely to be judged salient than others, and implies a relative rank exists on salient objects. The solution presented in this paper solves this more general problem that considers relative rank, and we propose data and metrics suitable to measuring success in a relative object saliency landscape. A novel deep learning solution is proposed based on a hierarchical representation of relative saliency and stage-wise refinement. We also show that the problem of salient object subitizing can be addressed with the same network, and our approach exceeds performance of any prior work across all metrics considered (both traditional and newly proposed).