The Secrets of Salient Object Segmentation

TL;DR

This paper critically evaluates existing salient object segmentation datasets, identifies biases, and introduces a new dataset with combined fixation and segmentation ground-truth to improve algorithm development.

Contribution

It reveals dataset design flaws, proposes a new high-quality dataset with combined ground-truth, and introduces a novel segmentation method bridging fixations and salient objects.

Findings

Identified biases in existing datasets

Created a new dataset with fixation and segmentation ground-truth

Achieved significant benchmark improvements

Abstract

In this paper we provide an extensive evaluation of fixation prediction and salient object segmentation algorithms as well as statistics of major datasets. Our analysis identifies serious design flaws of existing salient object benchmarks, called the dataset design bias, by over emphasizing the stereotypical concepts of saliency. The dataset design bias does not only create the discomforting disconnection between fixations and salient object segmentation, but also misleads the algorithm designing. Based on our analysis, we propose a new high quality dataset that offers both fixation and salient object segmentation ground-truth. With fixations and salient object being presented simultaneously, we are able to bridge the gap between fixations and salient objects, and propose a novel method for salient object segmentation. Finally, we report significant benchmark progress on three existing datasets of segmenting salient objects